91350 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Policy Research Report A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY CONCEPTS, DATA, AND THE TWIN GOALS © 2015 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 17 16 15 14 This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denomina- tions, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http://creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: World Bank. 2015. A Measured Approach to Ending Poverty and Boosting Shared Prosperity: Concepts, Data, and the Twin Goals. Policy Research Report. Washington, DC: World Bank. doi:10.1596/978-1-4648-0361-1. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third- party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permis- sion is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. ISBN (paper): 978-1-4648-0361-1 ISBN (electronic): 978-1-4648-0362-8 DOI: 10.1596/978-1-4648-0361-1 Cover image: © David S. Rose / Shutterstock.com. Used with permission; permission required for reuse. Cover design: Naylor Design Library of Congress Cataloging-in-Publication Data has been requested. This report reflects data available up to August 2014. Contents Foreword xi Acknowledgments xiii Abbreviations xv Overview 1 Evidence as the foundation for policy design 3 Ending global poverty 6 Boosting shared prosperity 9 Need for transformational policies 12 Alternative notions of poverty and shared prosperity 15 Challenges posed by uncertainty and downside risk 16 Monitoring poverty and shared prosperity 18 Complementary data for tracking poverty and shared prosperity across countries and over time 20 Concerted effort is needed to improve measurement methods and data 22 Notes 26 References 26 1. Defining and Assessing the Goal of Ending Poverty by 2030 29 A brief overview of global poverty measurement 30 Assessment of the global poverty target 40 Does the ending poverty target become more elusive when nearing success? 52 Poverty and shared prosperity 65 Notes 66 References 69 2. Understanding Shared Prosperity 73 The evolution of shared prosperity 74 Shared prosperity decoded 77 Tracking shared prosperity in practice 80 Can boosting shared prosperity help end global poverty? 97 Key conclusions on shared prosperity 105 Notes 106 References 107 v CONTENTS 3. The Twin Goals in a Broader Context 111 Welfare functions with poverty lines 114 Beyond the poverty line: Social welfare functions that care about everyone 119 Growth and social welfare 126 Going global: From country-level to global social welfare functions 130 Beyond income: Multidimensional social welfare functions 135 Notes 141 References 142 4. Uncertainty, Downside Risk, and the Goals 145 How uncertainty affects assessment of the goals 146 Sources of uncertainty about progress toward the goals 158 Uncertainty and the ability to assess progress toward the goals 178 Notes 180 References 180 5. National Profiles of Poverty and Shared Prosperity, Data, and Methods 187 Comparable household survey data for effective policy 187 Large data gaps and new technologies and methods 200 Notes 218 References 219 6. Global Profiles of Poverty and Shared Prosperity, Data, and Measurement Issues 225 Why census data are needed to count the poor 226 Purchasing power parity data: A unifying standard for measuring the poor 233 Measures of inflation and growth to align data to the same year 241 Notes 255 References 257 About the Team 263 Index 267 Boxes O.1 Structure of the report 3 O.2 Global poverty assessment since 1990 4 O.3 Why measure poverty in terms of income or consumption? 6 O.4 Frequently asked questions about the World Bank’s shared prosperity goal 9 vi CONTENTS O.5 Summary of the report’s key recommendations 25 1.1 Setting national poverty lines around the world 37 2.1 The World Bank’s early discussions of shared prosperity 75 2.2 Why 40 percent? 79 2.3 Measuring and tracking shared prosperity at the country level 86 2.4 The challenges of measuring and tracking shared prosperity at the global level 94 2.5 Does inequality affect income growth “equally”? 98 3.1 Social welfare functions articulate priorities across individuals 112 3.2 Where to draw the poverty line? 120 4.1 Uncertainty in forecasting economic growth 148 4.2 Modeling uncertainty in poverty projections 151 5.1 Shared prosperity is robust to measurement error in top earners 194 5.2 Household income and consumption surveys at the World Bank 202 5.3 Poverty maps and public policy: The case of Mexico 215 5.4 U.S. Census 2000: Short and long form details 217 6.1 Nonsampling error in population estimates 231 6.2 U.S. inflation and the international poverty line 243 6.3 Impact of spatial and temporal price differences on national poverty estimates: The controversial case of India 246 6.4 Lining up country surveys for aggregate poverty estimates 249 Figures BO.2.1 Number of surveys in PovcalNet over time 5 O.1 Global poverty projections are sensitive to underlying growth assumptions 8 O.2 The bottom 40 percent can encompass various income groups across countries 11 O.3 Shared prosperity has been correlated with average income growth 13 O.4 The goals appear more difficult to attain in the context of uncertainty and downside risk 17 1.1 Changing patterns of global poverty, 1981–2030 42 1.2 What does it take? Actual and required growth rates to achieve the aspirational scenario 51 1.3 Poverty reduction in the developing world, global measures 1980–2010 53 1.4 The effect of growth on poverty under the assumption of unchanged inequality 55 1.5 Declining sensitivity of poverty reduction would require ever- increasing growth 56 1.6 The trajectory of future poverty reduction may not be obviously linear 57 1.7 Heterogeneous subnational growth in Vietnam leads to slower national poverty reduction 61 vii CONTENTS 1.8 Poverty reduction in countries that have already achieved zero extreme poverty, 1820–2000 63 1.9 Poverty reduction in Thailand, 1981–2010 64 2.1 The bottom 40 percent in the United States, Brazil, and India, 2008 81 2.2 The bottom 40 percent can encompass various income groups across countries, circa 2009 82 2.3 The bottom 40 percent compared to the poor as defined by national poverty lines 84 2.4 Evolution of mean income or consumption of the bottom 40 percent and the overall population, 1980–2010 87 2.5 Shared prosperity in rural India at various levels of disaggregation, 2007/08–2009/10 89 2.6 Illustration of how the choice of data and time interval influence shared prosperity estimates 90 2.7 Moving averages provide more stable shared prosperity estimates 91 2.8 Shared prosperity, by country 92 2.9 Growth and changing shares of income 96 2.10 Shared prosperity and average income growth 99 2.11 The association of poverty reduction with overall income growth and shared prosperity 100 2.12 Twinning growth and shared prosperity to reach the 2030 extreme poverty goal 104 B3.1.1 Income distributions and social welfare functions 113 3.1 The headcount provides an incomplete picture of well-being below the poverty line 115 3.2 Welfare weights implied by different poverty measures 116 3.3 Different poverty measures fall with income, but tell different stories 117 B3.2.1 Weakly relative poverty lines, and global poverty based on weakly relative lines 120 3.4 Welfare weights implied by different social welfare functions 122 3.5 Social welfare increases with average income 125 3.6 High-end inequality and social welfare in the United States, 1950–2010 126 3.7 Growth and social welfare 128 3.8 Trends in global poverty measures, 1980–2030 131 3.9 Bottom 40 percent at home and in the world 133 3.10 Who in the world do the twin goals address? 134 3.11 E pluribus unum? Constructing multidimensional social welfare indicators 137 3.12 Welfare and per capita GDP 140 B4.1.1 The discrepancy between forecasted and actual growth since 2008 148 viii CONTENTS 4.1 Projecting global poverty headcount rates based on past growth rates, 2010–30 150 4.2 Drawing on past patterns and variation in growth to model uncertainty about future poverty rates, 2010–30 153 4.3 Uncertainty about the trajectory of poverty based on growth, 2010–30 153 4.4 Uncertainty about inequality contributes to further uncertainty about future poverty, 2010–30 155 4.5 The frequency of negative growth 156 4.6 Scenarios for global poverty under more frequent crises, 2010–30 160 4.7 The incidence of armed conflict in the world has declined since the early 1990s 165 4.8 Economic growth and poverty reduction are slower in Africa’s fragile states, 1981–2010 167 4.9 The share of the global poor living in fragile and conflict-affected states could double by 2030 168 4.10 Governance transitions are associated with temporary economic downturns and stronger long-term growth 169 4.11 Climate change incidence curves for rural India, 2040 175 5.1 Comparison of consumption measures resulting from different survey modules 191 5.2 Percentage difference between metro and nonmetro poverty in the United States, 1991–2002 199 5.3 Spatial variation in cost of basic needs, Bangladesh, 2010 200 5.4 Number of surveys in PovcalNet and reference years, 1978–2012 201 5.5 Comparison of imputation models for Sri Lanka’s poverty rate 210 5.6 Various imputation-based poverty estimates and actual poverty headcount between two cross-sections of household surveys in Vietnam, 1992/93 and 1997/98 211 5.7 Backward and forward imputation using combined data from two household surveys from Sri Lanka, 2006 and 2009 212 5.8 India consumption transition dynamics based on synthetic panel data estimated using the ELL approach, 2004–09 214 B6.1.1 Niger total fertility estimates, multiple data sources and methodologies, 1970–2012 231 6.1 Changing population projections and effects on poverty estimates, Bangladesh, 2005–15 232 6.2 Poverty over time in Bangladesh: Comparison of inflation indexes, 2000–10 247 B6.4.1 Illustration of open-ended lineup of survey into reference years 249 ix CONTENTS B6.4.2 Illustration of lineup into reference years between two surveys 250 6.3 Growth rates of survey consumption versus growth rates of national accounts consumption 252 6.4 Projection error in poverty estimates that use HFCE to scale up consumption, 2004–10 253 Maps 1.1 Poverty headcount at $1.25 a day, 2011 47 1.2 Poverty headcount at $1.25 a day, 2030 47 1.3 Increased spatial concentration of poverty in Vietnam, 1999 and 2009 60 B5.2.1 Most recent household consumption survey available in Africa 203 6.1 Census data coverage for the 2010 round 228 Tables O.1 Estimates of the percentage poor in 1993, based on three PPP indexes 21 1.1 Poverty in 2011 at $1.25 a day 2005 PPP 41 1.2 Ending global poverty 44 1.3 Alternative One: Projections based on countries’ experiences over the past 20 years 45 1.4 Alternative Two: Projections based on countries’ experiences over the past 10 years 46 1.5 Alternative Three: What do household surveys say? 48 1.6 Alternative Four: An aspirational scenario 49 1.7 Actual and required growth rates in the 10 countries contributing most to poverty in 2011 50 1.8 Actual and required growth rates in the 10 countries that will remain as principal contributors to global poverty in 2030 52 2.1 Reaching the 2030 extreme poverty goal in a world of higher shared prosperity 102 4.1 Increases in poverty before and during selected recessions 161 B5.1.1 Inequality and shared prosperity in Nepal, 1995–2011 194 5.1 Poverty and shared prosperity by quarter 196 6.1 Countries with outdated censuses 229 6.2 Poverty estimates from 1979 based on market and PPP-adjusted exchange rates 235 6.3 Estimates of the percentage of the poor in 1993 based on three PPP indexes 238 x Foreword It was roughly a year ago that the World Bank Group adopted two over- arching goals to guide its work. These goals seek to end extreme poverty by 2030 and to boost shared prosperity in every society. The goals are intended to give direction and galvanize action in the organization. It is the first time the World Bank will work toward a specific poverty target, and also the first time that it has given a call for all societies to strive for shared prosperity, which combines growth with equality. Following the formal announcement of the goals in April 2013, the World Bank has begun to assess what the new goals will mean for its work in different regions, and how the organization itself will need to change in response to the goals. This Policy Research Report contributes to this ongoing assessment, by focusing specifically on the data and measurement issues surrounding the goals. It lays out the conceptual underpinnings of the goals, assesses what reaching the goals will take, and reviews the data requirements for monitoring the goals. A clear understanding of the empirical basis of the goals is critical to ensuring success in achieving them. One of the key contributions of the report is to provide a detailed and comprehensive account of the data and processes needed to measure global poverty and shared prosperity. It dem- onstrates how many different data sources are needed—beyond household surveys—and highlights how sensitive measures of poverty and shared prosperity can be to changes in the underlying data. A major takeaway from the analysis is that much more support is needed to help build the capacity of data system architectures at the country level. While much work to date has focused either on poverty or on issues of inclusive growth, the report makes a strong case for viewing the two goals in unison. An emphasis on ending global poverty alone might lead the xi FOREWORD development community to focus almost exclusively on a few large coun- tries where poverty is high; the shared prosperity goal ensures attention is given to the least well off in all societies. The report helps highlight cases where the goals are intimately related, such that similar policies can be used to target both goals, as well as cases where reaching the goals may require different policies aimed at different groups of people. The report draws attention to the considerable challenge ahead of us if the twin goals are to be attained. While strong and sustained growth will be critical to meeting the goals, under reasonable growth rate assumptions extreme poverty is projected to remain well above the 3 percent target by 2030. The report shows how boosting shared prosperity, apart from being a desirable end in itself, can add considerable impetus to further poverty reduction toward the poverty target. Nevertheless, the projections suggest that reaching the global poverty target remains challenging and will require a departure from historical experience—of both growth and distributional effects and policies. This implies that achieving the goals will require concerted action and transformational policies that go well beyond “business as usual” practices. Adoption of the new goals will not only shape the agenda of the World Bank’s operational work in coming years, but will also give renewed impe- tus to research on these important issues. This report makes an important contribution to our understanding of global poverty and shared prosperity, but it is only a first step in this direction. I hope it will launch even more innovative research on the topic and inspire policy makers in different countries to develop and implement more effective policies toward these goals, for a better world. Kaushik Basu Senior Vice President and Chief Economist World Bank Group xii Acknowledgments This Policy Research Report was prepared by the Development Economics Research Group of the World Bank by a team led by Dean Jolliffe and Peter Lanjouw. The other authors of the report were Shaohua Chen, Aart Kraay, Christian Meyer, Mario Negre, Espen Prydz, Renos Vakis, and Kyla Wethli. Valuable contributions were made by Tatjana Kleineberg, Christoph Lakner, Prem Sangraula, Ilana Seff, and Liang Yu. Overall guidance of the report was provided by Kaushik Basu, Senior Vice President and Chief Economist, and Asli Demirgüç-Kunt, Director of Research, Development Economics. The team benefited from con- tinued engagement with the Global Poverty Working Group, including in particular Andrew Dabalen, Nobuo Yoshida, and Umar Serajuddin; and also with the poverty global practice, including in particular Ambar Narayan and Carolina Sanchez. The team also benefited from extended conversations with and feedback from Stefan Dercon, Patrick Gerland, John Gibson, and Martin Ravallion. For valuable feedback, the team thanks Pedro Alba, Inger Andersen, Antonella Bassani, Raka Banerjee, Kathleen Beegle, Jorge Calderon, Gero Carletto, Laurence Chandy, James Close, Hai-Anh Dang, Gabriel Demombynes, Makhtar Diop, Quy-Toan Do, Marianne Fay, Caroline Heider, Vivian Y. N. Hon, Olga Jonas, Talip Kilic, Philippe Le Houérou, Jeff Lecksell, Arianna Legovini, Jeff Lewis, Norman Loayza, Misha Lokshin, Aaditya Mattoo, Nicolas Mombrial, Cyrill Muller, Rose Mungai, Kyle Peters, Sanjay Pradhan, Martin Rama, Vijayendra Rao, Ana L. Revenga, David Robalino, Luis Serven, Klaus Tilmes, Anthony G. Toft, Laura Tuck, Roy Van der Weide, Joachim von Amsberg, Axel von Trotsenburg, Jan Walliser, and Ruslan Yemtsov. The team benefited from comments received at a seminar organized by Andrew Norton and xiii ACKNOWLEDGMENTS Kevin Watkins at the Overseas Development Institute in May 2014. The team also thanks the many others inside and outside the World Bank who provided comments. The World Bank’s Publishing and Knowledge Division coordinated the design, typesetting, printing, and dissemination of the report. Special thanks to Janice Tuten, Patricia Katayama, Susan Graham, and Andrés Meneses. The report was edited by Sandy Gain. The team also thanks Merrell Tuck-Primdahl and Ryan Hahn for their guidance on communi- cations strategy. xiv Abbreviations AIDS acquired immune deficiency syndrome CAPI computer-assisted personal interviewing CPI consumer price index ELL Elbers, Lanjouw, and Lanjouw FGT Foster-Greer-Thorbecke GDP gross domestic product GPS Global Positioning System HFCE household final consumption expenditures HIES Household Income and Expenditure Survey IAM integrated assessment model ICP International Comparison Program IMF International Monetary Fund IPCC Intergovernmental Panel on Climate Change MPI Multidimensional Poverty Index NSO national statistical office NSS National Sample Survey PPP purchasing power parity WDI World Development Indicators WPP World Population Prospects xv Overview The entrance to the World Bank’s headquarters in Washington, DC, is inscribed with the words “Our dream is a world free of poverty.” In pur- suit of this dream, in April 2013, World Bank President Jim Yong Kim announced to the international community two new goals to guide the World Bank’s work. First, it would seek to end global poverty, reducing the share of people living in extreme poverty to 3 percent of the global population by 2030. Second, it would seek to boost shared prosperity, understood as increasing the average incomes of the bottom 40 percent of the population in each country. The accompanying narrative emphasized that both goals should be attained in a sustainable and inclusive manner, ensuring that today’s development is not reversed tomorrow and does not compromise the planet’s future, or that of subsequent generations. The adoption of these two goals marks a significant shift for the World Bank. Although poverty reduction has been a mainstay of its work for decades, the World Bank has now, for the first time, committed to a spe- cific poverty reduction target to guide its work. Similarly, the goal to boost shared prosperity gives more explicit attention to inclusive growth than has been the case in the past and paves the way for a focus on inequality, not only of opportunity but also of final outcomes. Prosperity also needs to be shared across individuals over time, requiring forms of sustainable development that fully account for environmental degradation and natu- ral resource depletion as well as, crucially, their close interrelation with poverty. Articulation of and commitment to global development goals can help build momentum toward their achievement. The goals seek to provide the global development community with a unified sense of purpose and to 1 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY galvanize action around clear and easily communicable objectives. Beyond this motivational function, however, the goals also have a tangible link to the manner in which the World Bank and other development agencies conduct their operations. Although the World Bank alone does not have the capacity to realize these global goals, it has pledged to place them front and center in the institution’s work going forward. The indicators associated with the goals will also be used by other development agencies to target programs and allocate funding. The two new goals provide a new context for policy assessment. They provide a framework in which to evaluate policies and their potential con- tribution to poverty reduction and inclusive growth. Assessing progress toward the goals (or lack thereof) provides a means to identify gaps and prioritize actions. In this way, the goals not only help regional and inter- national donors to target available resources, but also inform national gov- ernments in their efforts to reduce poverty. However, while assessment of progress toward the goals will provide a benchmark for the World Bank’s dialogue with countries about poverty reduction, the precise way in which those priorities are set and achieved should be determined at the country level, according to countries’ own policies and circumstances. How progress is measured will matter. The World Bank’s choice of indicators reflects particular institutional priorities, prompted by criteria that balance precision and conceptual coherence with ease of communica- tion and global comparability. Alternative measures may provide differ- ent insights. By offering a fuller exploration and exposition of the global poverty and shared prosperity goals, this report seeks to provide a richer basis from which individual countries can choose measures that are most relevant to their circumstances. This report goes beyond motivating the importance of the new goals, to focusing squarely on issues of measurement and data. The objective of the report is to articulate what measuring the poverty and shared prosperity goals entails and to identify areas where improvements in data are needed to assess and monitor them. This discussion is fundamental to achieve- ment of the goals. Action to reduce poverty and boost shared prosperity would be greatly impaired without the ability to credibly and consistently measure progress. The chapters that follow lay out the conceptual under- pinnings of the World Bank’s two goals and assess what reaching them will require; they discuss the relative strengths and weaknesses of the goals by contrasting them with alternative indicators; and they propose empirical approaches to tracking their progress (box O.1). 2 OVERVIEW Box O.1 Structure of the report This Policy Research Report is structured in three confidence in achieving the goals and indeed their parts, mirroring the three broad aims of the report. very attainment are sensitive to assumptions about The first part provides a general overview of the the patterns of economic growth and the occur- conceptual underpinnings of the two goals and rence of extraordinary shocks. their assessment. Chapter 1 describes the World Finally, while data and measurement issues are Bank’s approach to poverty measurement and discussed throughout, the third part of the report assesses what achievement of the poverty goal will specifically addresses issues related to the empirical require. Chapter 2 turns to the shared prosperity monitoring of the goals in greater technical detail. goal, demonstrating how the goal can be evalu- Chapter 5 discusses the use of household survey ated and highlighting some of the challenges of data in measuring global poverty and shared pros- interpretation. perity, highlighting some of the challenges faced The second part of the report places the World in raising the frequency and timeliness of global Bank’s two goals in a wider context. Chapter 3 poverty estimates. Although household surveys places the global poverty and shared prosperity are necessary inputs to the measurement of global goals in a broader framework of poverty and welfare poverty and shared prosperity, they are not suf- analysis. It shows how the World Bank’s choices of ficient. Chapter 6 thus turns to some of the key measures are two options from an array of possible complementary data—population data, purchas- indicators, each with different features that provide ing power parity (PPP) indexes that control for the different insights. Chapter 4 discusses poverty pro- differences in the cost of living across countries, jections in the context of uncertainty about eco- and growth and inflation data—that are needed to nomic growth and large or unusual shocks, which support the World Bank’s poverty and prosperity could pose downside risk to achieving the goals estimates. The discussion on accounting for differ- and are often not adequately captured by standard ences in prices across countries with PPP indexes economic models. Current debates around climate is particularly extensive, primarily because these change and sustainability receive explicit attention data have significant implications for global pov- in this framework. The chapter demonstrates how erty estimates. Evidence as the foundation for policy design Concerns around data and measurement are often overshadowed in debates about the fundamental determinants of development and the role of policy. This report argues for a different perspective—one that acknowledges the role evidence plays in understanding structural change and the design of policy and appreciates the importance of evidence in evaluating and improving policies over time. Economists rely on the availability of con- sistent and reliable data not only to motivate and assess economic theory, but also to monitor and evaluate economic policies in practice—and this is as important for poverty reduction as for other areas of economics. As the 3 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY eminent Indian statistician Prasanta Chandra Mahalanobis once declared, “. . . statistics is an applied science and . . . its chief object is to help in solv- ing practical problems. Poverty is the most basic problem of the country, and statistics must help in solving this problem” (Mahalanobis 1963). Far from being an issue of secondary importance, data and measure- ment are pivotal to the assessment of the World Bank’s new goals and, thereby, their achievement. To assess progress toward the goals, it is necessary to have a clear understanding of how progress is defined and measured. Without a clear understanding of the goals’ meaning and know- ing how to measure progress, what would be the basis for selectivity and prioritization? And how would lessons be learned from past experience? This report will argue that improved data infrastructure—consisting of many elements, including more attention to measurement methods and the collection of more and better survey data as well as complementary population and price data—is critical to ensure that progress toward the goals can be measured and policies to help achieve them can be identified and prioritized. Although the availability of poverty data has increased substantially in the past few decades, infrequent or unreliable data continue to pose a challenge to global poverty assessment (box O.2). Box O.2 Global poverty assessment since 1990 While poverty reduction has been a mainstay of Perhaps more important is the impact this the World Bank’s mission for decades, the mea- increased ability to measure poverty has had on surement of global poverty has at times lagged poverty reduction efforts. Poverty assessments, behind ambitions to reduce it. The 1990 World drawing on country-level poverty data, inform Development Report was an important milestone countries’ understandings of the plight of their in global poverty assessment, providing one of the citizens and help countries to shape policies accord- fi rst comprehensive cross-country databases on ingly. Such analyses have become increasingly com- poverty and a concerted effort to articulate what mon and detailed alongside the expansion of data. was needed to improve the measurement of pov- At the global level, improved data have supported erty. That effort was based on single household international efforts to reduce poverty, includ- surveys from 22 countries. The World Bank now ing by providing the basis for the Millennium has access to more than 1,000 surveys from 1981 to Development Goal aimed at halving global poverty 2011 (figure BO2.1), covering nearly all developing between 1990 and 2015. countries—making national poverty assessments However, although encouraging progress has possible in most countries. been made in improving the quantity of household (continued) 4 OVERVIEW Box O.2 continued surveys needed to measure poverty, this report a chronic or transitory condition), or the deter- details the remaining challenges with the frequency minants of poverty. Improved poverty analysis and quality of data. Although most countries now requires more than just an increase in the number have national poverty assessments, the global devel- of surveys available. Concerted efforts are required opment community does not yet have the consis- to improve the capacity for data collection at the tent and frequent data needed to understand fully country level to produce not just more, but also the nature of poverty in countries, the evolution of better-quality poverty data. poverty over time (and whether poverty is largely Figure BO.2.1 Number of surveys in PovcalNet over time 160 PovcalNet reference year 140 120 Three-year moving total Number of surveys 100 80 60 Surveys per year 40 20 0 78 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20 20 20 20 Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: It is quite common for there to be a delay of some months between when a survey is collected, when it is published, and when it becomes available in PovcalNet. The decline between 2010 and 2011 illustrated in the figure therefore reflects the fact that many surveys collected in 2011 are not yet available in PovcalNet, rather than a substantial decline in the number of surveys collected in 2011. At the same time, discussion of improvements in data that are needed to measure poverty and shared prosperity consistently across countries should not ignore the progress in data measurement, access, and quality that has been achieved in recent decades. The purpose of this report is to identify areas where further improvements can build on progress that has already been made and highlight particular areas where further progress is needed. 5 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Ending global poverty In the past few decades, substantial progress has been made in reducing global poverty. The World Bank assesses poverty by the number of people whose income or consumption falls below a given threshold (box O.3). Between 1990 and 2011, the number of people living in extreme poverty has halved, to around one billion people, or 14.5 percent of the world’s population (17.0 percent of the developing world’s population). While this progress is encouraging, the fact that so many people remain poor is sober- ing. To estimate the number of people living in extreme poverty, the World Bank currently uses an international poverty line of $1.25 a day, in 2005 prices—a poverty line that corresponds to an average of the national pov- erty lines of the 15 poorest developing countries.1 That more than a billion people in 2011 eked out a living on such a low threshold living standard makes the need to increase efforts to reduce global poverty self-evident. Why set the global target for poverty reduction to 3 percent of the world’s population by 2030? The 3 percent target derives from conceptual and Box O.3 Why measure poverty in terms of income or consumption? The measurement of poverty using income or con- to aggregate across different dimensions to con- sumption has a long tradition, although consump- struct a multidimensional measure of welfare and, tion is usually the preferred indicator in developing if so, how to do so in a way that is conceptually countries. Consumption is typically assumed to be sound and readily interpretable. Stopping short less volatile than income and is thus often seen as a of attempting to construct an explicitly multidi- better measure of current living standards. From a mensional measure of poverty does not, however, practical perspective, consumption is usually more mean that comparisons of deprivation along vari- easily and accurately measured than income in coun- ous dimensions are not possible. World Bank pov- tries with relatively low levels of participation in erty assessments, carried out at the country level, formal labor markets. However, in countries where routinely look not only at consumption poverty, income is the only available indicator of economic but also at deprivation along other dimensions. welfare, particularly in Latin America, measuring Similarly, global consumption poverty can and poverty on the basis of income data is the norm. should be examined alongside deprivations in all The choice to measure poverty in terms of other relevant dimensions. Importantly, households income or consumption should also be distin- may experience multiple forms of hardship simulta- guished from multidimensional poverty mea- neously and, where this happens, policy responses sures. Although there is widespread consensus must recognize and address these joint depriva- that poverty is a multidimensional phenomenon, tions. This discussion is taken up in greater detail there is much less consensus on whether it is useful in chapters 1, 2, and 3. 6 OVERVIEW empirical considerations. Conceptually, it may be desirable to set a target to eliminate global poverty altogether. However, a global goal of zero poverty would require eliminating poverty in each and every country. Poverty in some countries remains deep and widespread, and it is simply not realistic to expect to be able to eliminate poverty in these countries by 2030. It is also the case that at any moment in time there is likely to be some churning taking place in which some people, possibly for reasons beyond their control, fall into poverty, even if only temporarily. It is thus practical to set a global target close to zero, but which allows for some heterogeneity at the country level. Empirically, simple back-of-the-envelope simulations can be con- ducted to assess the plausibility of the goal to end poverty by 2030. When such simulations are based on highly stylized and rather optimistic assumptions—such as stable and continuous annual growth rates in con- sumption per capita of at least 4 percent in all developing countries and an unchanging distribution of income—then a global poverty rate of 3 percent is achievable.2 Such analysis suggests that the World Bank’s dream of ending global poverty by 2030 is a highly aspirational objective, but is not entirely beyond reach with concerted efforts and commitment from individual countries as well as the international development community. To say that the global poverty goal could be reached with concerted effort, however, is not to say that doing so would be easy. Although per capita growth of 4 percent in each country is roughly equivalent to the average for developing countries as a whole from 2000 to 2010, assuming that all countries could consistently grow at this rate is highly implausible. In the past three decades, such growth rates have been far from common (figure O.1, panel a). If developing countries were instead to grow at their respective annualized growth rates of the past 20 years, global poverty would remain at around 6.8 percent of the world population by 2030, a considerable distance from the 3 percent target (figure O.1, panel b). Chapter 1 sets out a series of alternative growth simulations and assesses the likely impact of each on global poverty. Together, these combine to emphasize that the World Bank’s goal to reduce poverty to 3 percent of the population by 2030, while not impossible, is certainly ambitious. Not only would achievement of the global poverty goal require strong economic growth, there is some evidence that the poverty target may become more difficult to reach as it becomes closer. Although there has been a striking linearity in the decline of the global poverty headcount since the early 1980s, the future path toward the 3 percent target may entail a significant tapering off of progress. One reason is that, although it 7 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure O.1 Global poverty projections are sensitive to underlying growth assumptions a. Historical growth rates across countries b. Global poverty projections to 2030 40 0.10 35 Poverty reduction Global poverty headcount (percent) 30 since 1990 25 Density Countries 20 grow at their 0.05 respective averages over 15 past 2 decades 10 All countries grow at 4% per year 5 3% poverty target 0 0 −20 −10 0 10 20 1990 1998 2006 2014 2022 2030 GDP per capita growth (percent) Source: Based on data from the World Bank World Development Indicators database, panel a, and the World Bank PovcalNet database, panel b. Note: Panel a shows the frequency of different annual per capita gross domestic product (GDP) growth rates for 129 countries between 1980 and 2010. The dashed orange line denotes per capita growth of 4 percent. may initially be possible to reach many poor people through broad-based economic growth that generates more and better-paid jobs, as poverty declines it may be relatively more difficult to reduce poverty in hard-to- reach geographic pockets or among population groups that are somehow excluded from participation in the broader economic currents. In countries experiencing conflict, it may be particularly difficult to reach populations in affected areas. In some cases, the poor may be veritably trapped in pov- erty because of failures in credit, land, or other key markets, or because low levels of education, skills, or health prevent them from availing themselves of new opportunities proffered by a general expansion of economic activ- ity. Such factors, which contribute to unevenness in the rate of poverty reduction in countries, can result in a declining responsiveness of poverty reduction to a given rate of aggregate growth over time. On the other hand, experience from currently high-income countries indicate that the trajec- tory of poverty decline does not inevitably taper off as it approaches zero. Concerted efforts by policy makers can help to maintain progress. These 8 OVERVIEW efforts are likely to involve a focus not only on average income growth but particularly on raising the incomes of the poor. Boosting shared prosperity The World Bank’s second goal, of boosting shared prosperity, places increased focus on the least well-off in society. Discussion of inclusive growth is not new. However, although there is an extensive literature emphasizing the importance of thinking about inclusion of the poorest in society in defining goals for development, until now there has not been agreement on a single summary indicator. The World Bank’s new shared prosperity goal—to boost the incomes of the bottom 40 percent of the population—provides a measure of inclusive growth (box O.4). Box O.4 Frequently asked questions about the World Bank’s shared prosperity goal In April 2013, World Bank President Jim Yong How is shared prosperity different from aver- Kim announced a global goal of promoting shared age income as a measure of development progress? prosperity. This new goal often confronts some The shared prosperity measure places explicit common questions: emphasis on the least well-off in society, focus- ing on the bottom 40 percent. In addition, unlike How will shared prosperity be measured? Boosting growth in gross domestic product per capita, shared prosperity is understood by the World Bank assessed from national accounts data, the shared to mean fostering the well-being of the bottom 40 prosperity indicator is assessed from household percent of the population in every country. This survey data. The two measures are thus not will be assessed by measuring the income or con- directly comparable. sumption growth of the bottom 40 percent of the population in each country over time. Is this an inequality measure? Tracking the income What does the “shared” in shared prosperity mean? growth of the bottom 40 percent of the popula- Shared refers to the extent to which the bottom 40 tion is not sufficient to gain insight into changes percent of the population takes part in and benefits in inequality. However, by comparing the shared from the process of economic development. prosperity measure with a survey-based measure of average income or consumption, or of income What is “good” shared prosperity performance? The growth of the top 60 percent of the population, it goal does not make a normative statement about is easy to use the shared prosperity measure to what defines good shared prosperity: the higher the learn about the evolution of inequality in coun- growth rate of the average incomes (or consumption) tries over time. The details of this are explored of the bottom 40 percent of the population, the better. in chapter 2. 9 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY One way to think about the World Bank’s new shared prosperity goal is as an alternative to average income as the benchmark of development progress. Instead of assessing and measuring economic development in terms of the overall average growth in a country, the shared prosperity goal places emphasis on the bottom 40 percent of the population. In other words, good progress is judged to occur not merely when an economy is growing, but, more spe- cifically, when that growth is reaching the least well-off in society. Thus, the shared prosperity goal seeks to increase sensitivity to distributional issues, shift- ing the common understanding of development progress away from average per capita income and emphasizing that good growth should benefit the least well-off in society. This discussion is relevant in developed as well as develop- ing countries, since, notwithstanding the substantial progress that has been made in reducing absolute poverty in recent decades, in many middle- and high-income countries there is a concern that the relatively poor are being left behind. When poverty is viewed as an inability to participate and prosper in society, it remains a pervasive problem, even in developed countries. Unlike the World Bank’s global poverty goal, the shared prosperity goal is a country-specific goal, which does not have an explicit endpoint. It is unbounded, in that boosting shared prosperity requires a positive growth rate for the average incomes of the bottom 40 percent of the population, but there is no target (or limit) for what that growth rate should be. The shared prosperity indicator is thus similar to measures of average income, such as growth in gross domestic product (GDP) per capita, in its expres- sion (as a simple growth rate over time) and in how it is evaluated (more growth is better, without a specific target rate of growth in each country). However, the shared prosperity indicator has substantially different measurement requirements. Unlike GDP per capita, which is measured from national accounts data, the shared prosperity indicator needs to be measured from household survey data (which are also used for poverty assessment). This is because national accounts data only provide aggre- gated information on economic performance, not the disaggregated infor- mation on people living on different levels of income or consumption, which is needed to measure the income of the poorest 40 percent of society. Unlike national accounts data, which are produced on an annual basis in a relatively standard way, the frequency and quality of household survey data are heterogeneous, raising substantial challenges for cross-country comparisons. Although it does not seek to provide all the answers, this report offers a detailed discussion of these challenges and points to some possible improvements. 10 OVERVIEW Figure O.2 The bottom 40 percent can encompass various income groups across countries Russian Federation Thailand Tunisia Turkey Chile Armenia Costa Rica Peru Kyrgyz Republic Brazil China South Africa Honduras Ethiopia India Lao PDR Bangladesh Angola Mali 0 20 40 60 80 100 Population share (percent) Extreme poor (less than $1.25 a day) Moderate poor ($1.25 to $4 a day) Vulnerable ($4 to $10 a day) Middle class and rich (more than $10 a day) Source: Based on data for latest year available from the World Bank PovcalNet database (accessed August 2014). Note: The vertical line in the figure illustrates the bottom 40 percent of each country’s population—that is, the group that would be the focus of efforts to boost shared prosperity. The groups in the figure are the extreme poor, as defined by the World Bank’s international poverty line; the moderate poor, who live on between $1.25 and $4 a day; the vulnerable, who live on between $4 and $10 a day; and the middle class and rich, who live on more than $10 a day—all measured at 2005 constant purchasing power parity (PPP). The concept of people living on between $4 and $10 a day being considered vulnerable is based on evidence that a considerable share of households above a given poverty line is usually vulnerable to falling below that line over time. See Ferreira and others (2012), López-Calva and Ortiz-Juarez (2014), and Birdsall, Lustig, and Meyer (2014). Chapter 2 illustrates how the composition of the bottom 40 percent is very different across countries. In low- and lower-middle-income countries, there will likely be significant overlap between those living in extreme pov- erty and the bottom 40 percent of the population (figure O.2). Tracking shared prosperity can thus reinforce poverty reduction efforts in these 11 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY countries. By contrast, a substantial proportion of the bottom 40 percent of the population in upper-middle-income countries is likely to be nonpoor according to the global $1.25 a day standard. In these countries, tracking shared prosperity can bring attention to those who may not be covered by poverty policies but who might otherwise be relatively disadvantaged. The shared prosperity goal is not an inequality goal in and of itself. Measuring the income growth of the bottom 40 percent of the population provides no information on how that compares with the income growth of the rest of the population. However, an impression of inequality can easily be obtained by comparing the shared prosperity indicator with mean income growth (or income growth of the top 60 percent of the population). In this sense, the shared prosperity measure implicitly places emphasis on changes in inequality in society. It is noteworthy that while the World Bank has a fairly long-standing record of discussing the policies needed to create equality of opportunity, the shared prosperity entry point into the discussion of inequality is through an emphasis on equality of outcomes (in this case, in people’s relative incomes). This is a rather novel perspective for the World Bank. Need for transformational policies The analysis in chapter 1 highlights the critical role of continued growth in helping to reduce poverty. In all the simulations presented in chapter 1, growth contributes to poverty reduction, and the extent of the contribu- tion increases when the assumed underlying growth rates are higher. The important role of growth is also evident in backward-looking assessments. For example, growth in average incomes has historically been strongly correlated with growth in the incomes of the bottom 40 percent of the population (figure O.3). Put differently, and as shown in chapter 2, analysis of the relative contributions from boosting overall growth (increasing the size of the pie) and reducing inequality (increasing the poor’s slice of the pie) suggests that increased growth has played a more prominent role in boosting shared prosperity in the past. Achieving the World Bank’s goals will therefore require strong and sustained growth in developing countries. The analysis in chapter 1 also suggests, however, that continued growth in line with what has been experienced in recent decades will not be suffi- cient to end poverty. Under a variety of plausible growth rate assumptions, extreme poverty is projected to remain well above the 3 percent target by 12 OVERVIEW Figure O.3 Shared prosperity has been correlated with average income growth 15 Annualized income or consumption growth rate Bolivia of the bottom 40 percent (percent) 10 Cambodia Tanzania Colombia China 5 South Africa Ecuador Mali India Dominican Republic 0 Nigeria Serbia Madagascar –5 –10 –10 –5 0 5 10 15 Annualized income or consumption growth rate of the total population (percent) Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: Growth rates in shared prosperity are calculated as annualized growth rates in per capita income or consumption expenditure over the period of circa 2006–11 (all survey based). See note to figure 2.8 for further explanations. 2030; achieving the 3 percent poverty goal on the basis of growth alone would require national growth rates well above historical precedents. This suggests that achieving the poverty goal will require concerted action and transformational policies that go well beyond “business as usual” practices. Ending global poverty and boosting shared prosperity will require not just a focus on overall levels of growth, but particular attention to the nature and patterns of growth. Although the incomes of the poorest have tended to be correlated with average income growth in the past, there are also notable exceptions, where overall growth has not translated to effective poverty reduction or has taken place alongside increased inequality. This suggests that it is not just growth, but also the type of growth (growth that benefits the poor) that will be important to achieving the World Bank’s goals. Although this report does not set out detailed policy prescriptions for poverty reduction, given its primary focus on data and measurement issues, it is worth recognizing past analyses that emphasized the importance of different types of growth and the relative impact they have on poverty.3 In particular, growth that is widely shared and increases the returns to assets 13 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY held by the poor (especially the returns to their labor, but also to other assets, such as land holdings) is the most likely to translate into effective poverty reduction.4 Improved access to education, health, and capital can be critical elements in enhancing the returns to the poor’s assets. The World Bank’s second goal, of boosting shared prosperity, may also help in shifting focus toward the poor. Simulations in chapter 2 show that if growth in developing countries up to 2030 were to follow the same distributional pattern as growth of the respective regional leaders in terms of pro-poor performance over the past decade, this could contribute to reaching the global poverty goal. Another simulation considers increasing growth of the bottom 40 percent by 1 and then 2 percentage points more than the 10-year historical mean growth rate for each country (and cor- respondingly reducing growth of the top 60 percent to leave the overall mean growth rate unchanged). With some differences across regions, both of these simulations illustrate how boosting shared prosperity can add con- siderable impetus to further poverty reduction toward the poverty target.5 The poverty target is reached a few years ahead of 2030 in the simula- tion where income or consumption growth of the bottom 40 percent is increased by 2 percentage points more than the mean. These findings point to an important complementarity between the two goals, and it is in this sense that the global poverty and shared prosperity goals can be considered “twin” goals: achieving progress in both goals will require efforts on both fronts. However, the projections suggest that reaching the global poverty target remains challenging, suggesting that a departure from the historical experience—of both growth and distributional effects and policies—will be needed if the World Bank’s goals are to be met. As discussed above, in announcing its new goals, the World Bank stressed that the path toward them must be environmentally, socially, and economically sustainable over time. Thus, while boosting growth will be crucial to meeting both the global poverty and shared prosperity goals, the extent to which development trajectories compromise future growth and sustainable development will be important. There is a substantial literature demonstrating the importance of natural resources for sustainable eco- nomic development, not only because the poor often rely disproportion- ately on access to natural resources to meet their immediate needs, but also because degradation of natural resources can have profound impacts on the health and livelihoods of the poor as well as future growth prospects.6 This underscores the importance of developing policies that achieve growth in a sustainable way that does not undermine future progress. 14 OVERVIEW Alternative notions of poverty and shared prosperity Although the World Bank has selected two indicators to measure progress toward its goals—the number of people living in extreme poverty, as a share of the world’s population, and growth in the incomes of the bot- tom 40 percent of the population in each country—these are not the only possible indicators to measure progress in these important domains. The measures adopted by the World Bank and its development partners reflect particular institutional priorities, but individual countries may have dif- ferent priorities and may choose to emphasize other specific distributional features. The World Bank’s articulation of the goals and the choice of measures to monitor and assess these goals derive from its global vantage point. Thus, the incidence of poverty in the world as a whole has been set as the poverty target, rather than a country-specific target tailored to each indi- vidual country’s circumstances. Similarly, there is a clear intention not just to track shared prosperity in each country, but also to compare progress across countries. Individual countries may engage with the ideas behind these goals with a perspective that is more country specific. To that end, it is important to recognize that the specific goals defined by the World Bank can be seen as two particular applications of a whole class of approaches. An important element in clarifying and understanding the World Bank’s two new goals therefore comes from appreciating where the goals fit relative to a spectrum of alternatives, with each alternative providing different insight into social welfare. Ideally, a rich understanding of poverty and distribu- tional issues would be based on assessment of many or all of these measures. The scope for differences in priorities is clear in the context of poverty measurement. In contrast to the World Bank’s global poverty threshold, based on a single global poverty line, national governments attach priority to poverty thresholds that are more relevant to their particular countries, as evidenced by large differences across countries in national poverty lines. Countries may also prefer to go beyond the headcount measure of poverty to consider poverty measures that capture the depth and severity of poverty alongside the incidence of poverty. Similarly, the potential for different priorities can be seen in the context of the shared prosperity measure: the consumption or income share of the bottom 40 percent is just one of many measures of how equitably or inequitably income is distributed across individuals in a country, and different inequality measures imply different priorities over individuals at different points in the income distribution. 15 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Chapter 3 uses welfare functions as a tool of analysis to set the twin goals in this broader context. Economists have long used social welfare functions to capture societal preferences over how income is distributed across individuals in a society. The chapter first discusses social welfare functions that in some way distinguish the poor from the nonpoor, as the World Bank’s global poverty goal does. The choice within this group of functions is essentially about where and how to set the line distinguish- ing the poor from the nonpoor and whether to assess not just whether an individual is poor, but also how poor the individual is. A second group of welfare functions does not distinguish between the poor and the nonpoor. Instead, it considers the well-being of everyone in the income distribution, but places different weights on different groups of people, as is the case with the World Bank’s shared prosperity goal. The main choice in this group of functions is about what weight to place on individuals at different parts of the income distribution. Challenges posed by uncertainty and downside risk Chapter 4 discusses the World Bank’s goals in the context of uncertainty and downside risk. The scenarios for global poverty and shared prosperity presented in chapters 1 and 2 show how projections for the World Bank’s goals are sensitive to underlying assumptions, in particular about future growth. There is considerable uncertainty about the future trajectory and distributional nature of growth in developing countries, in turn implying uncertainty about trajectories for global extreme poverty and shared pros- perity. Although economic models can to some extent capture uncertainty in future projections through analysis of past experience, they are inher- ently limited in cases where the future may systematically diverge from the past. For example, recent debates around climate change have emphasized the difficulty in anticipating and predicting the economic consequences of continued rapid rises in global temperatures. Chapter 4 incorporates uncertainty into projections of global poverty and discusses the potential impacts of a selection of sources of downside risk. When projections are based on average growth rates from the past, the projections can be highly sensitive to the period on which the average is based (figure O.4, panel a; chapter 1). Furthermore, incorporating uncer- tainty and downside risk into projections of global poverty demonstrates not only that confidence in reaching the goals is diminished in the presence 16 OVERVIEW Figure O.4 The goals appear more difficult to attain in the context of uncertainty and downside risk a. Projections based on average growth rates in b. Projections based on varied growth rates in the past the past 18 18 Global poverty headcount (percent) 16 Global poverty headcount (percent) 16 14 14 12 12 10 10 8 8 6 6 4 3% target 4 3% target 2 2 0 0 30 30 10 12 14 16 18 20 22 24 26 28 10 12 14 16 18 20 22 24 26 28 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 1981–91 1991–2001 2001–11 Source: Based on data from the World Bank PovcalNet database. Note: Panel b shows probabilistic scenarios based on random draws from past variation in growth rates between 2000 and 2010. While the median projection from this exercise is for the poverty headcount to decline to 5.1 percent by 2030, panel b shows it could also be as high as 7.1 percent. of uncertainty, but that the goals can appear even more difficult to attain. If the growth assumptions underlying poverty projections for individual countries are allowed to fluctuate in line with patterns observed in the past (rather than assuming countries consistently grow at some average histori- cal growth rate), the 3 percent poverty goal remains difficult to reach, even under relatively optimistic assumptions (figure O.4, panel b). Similarly, the trajectory of future poverty is much more uncertain when the incidence (distribution) of growth is allowed to vary. Although difficult to model, factors that could influence the sustainability of growth are also likely to make the World Bank goals difficult to reach. This reinforces the message from chapter 1 that the 3 percent poverty target is not easily reached. What factors are likely to contribute to significant variation in the pace and incidence of growth in the future? There are many factors that will affect future growth, many of which are accounted for in projections of global poverty and shared prosperity by rooting the projections in the patterns of growth observed in the past. The discussion in chapter 4 thus focuses on exceptional economic and financial crises; fragility, political 17 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY instability, and conflict; climate change; and global pandemics because these are factors that could have profound implications for poverty and shared prosperity. However, the likelihood and magnitude of their possible impacts are not well understood. The World Bank’s goal to end poverty is a global goal, with its target expressed as a share of the world’s population. The goal is not articulated as one that aims to reduce global poverty to 3 percent in each and every coun- try by 2030. However, although it might be possible to achieve the goal through rapid poverty reduction in some but not all countries (especially if poverty reduction occurred in some of the most populous countries), it is clear that a poverty rate that remained very high in some countries by 2030 would in some sense negate the spirit in which the World Bank’s goal was articulated. In this respect the discussion of Africa in chapter 4 raises particular cause for concern. Notwithstanding the significant reduction in the incidence of armed conflict globally since the 1990s, almost half the countries in Africa are defined as being fragile states whose growth performance and poverty reduction have been lagging behind that of non- fragile states in the region. Sub-Saharan Africa is projected to be the region with the highest remaining prevalence of extreme poverty in 2030 in all the projections presented in chapter 1. Furthermore, several studies have found that the impacts on growth and poverty from temperature increases associated with climate change are likely to be particularly pronounced in Sub-Saharan Africa. Targeted support for the region will be needed to ensure that global poverty reduction leaves no country behind. In terms of the emphasis on the World Bank’s goals being achieved in a sustainable manner, the threats posed by climate change may be the most prominent source of uncertainty about future sustainability. Overall, cli- mate change will likely have a limited aggregate impact on extreme poverty and shared prosperity by 2030. However, it is expected to affect the long- term sustainability of development progress beyond 2030, particularly in the latter part of the century when more catastrophic climate change would likely occur under current emissions scenarios. Monitoring poverty and shared prosperity Having stressed the importance of data and measurement issues in assess- ing progress toward the twin goals, what, then, are the most important measurement and data requirements? Monitoring the goals requires many inputs, but comparable household survey data are the critical element. 18 OVERVIEW Household surveys provide information on people’s consumption or income—key variables in the World Bank’s approach to assessing poverty. Chapter 5 discusses the measurement of poverty and shared prosperity from household survey data in detail. Although the number of household surveys has increased in countries around the world and the quantity and quality of survey data in some developing countries are excellent, overall the fre- quency and quality of household survey data are highly variable and there are issues of consistency and comparability across and within countries. Heterogeneity across countries in the measure of consumption or income used to assess poverty and shared prosperity is not necessarily an indication of poor quality. In many cases, heterogeneity in surveys across countries may reflect differences that are important to take into account to assess poverty accurately at the country level, such as customization of surveys to take account of differences in living conditions in low- versus middle-income countries or differences in the types of food that are locally available. When tailoring questionnaires to local conditions produces data that are more useful to the country, then this sort of difference in the questionnaire design is desirable because it helps to produce policies that can be more effective in reducing poverty. Heterogeneity in household surveys from the same country over time, however, may be more problematic and is often the result of happenstance rather than intentional design. For example, changes to questionnaires often reflect changes in funding sources (with surveys altered to reflect donors’ interests) or simply changes in the personnel of data management teams. Changes to survey questionnaires can have substantial impacts on poverty estimates and make it difficult to answer simple questions, such as whether poverty has declined over time. For example, Beegle and others (2012) implemented an experiment where different consumption question- naires were randomly assigned to different subsamples in Tanzania. They found large variation in measured consumption and poverty estimates that were induced simply by the differences in how questions were asked. As one example, their results show that changing the recall period from one week to two weeks (leaving everything else the same) had the effect of increasing the estimated poverty headcount from 55 percent to 63 percent. Changes to questionnaires are often based on the notion that they will improve the informational content of the data, but typically little weight is placed on the cost imposed by creating noncomparable data. Effective policies to reduce poverty need to be informed not only by the overall level of poverty, but also by the geographic profile of poverty. Even when consumption data are consistently collected, it is impossible to 19 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY accurately estimate the subnational profile of poverty if spatial differences in the cost of living within the country are not taken into account. For example, Jolliffe, Datt, and Sharma (2004) estimate region-specific pov- erty lines in Egypt and show that accounting for spatial differences in the cost of living had the effect of increasing the poverty rate in capital cities by more than 150 percent. Without adjusting for cost-of-living differences, policy makers would have assumed that poverty rates in metropolitan areas were about a third of what they were in the rest of the country. Thus, to inform policy empirically and design policy effectively, it is critical for sur- veys to be not just temporally comparable, but spatially comparable as well. In the past, poverty has only been periodically assessed, and progress in reducing poverty was based only on these relatively infrequent measures. However, monitoring progress toward the World Bank’s goals in a credible and consistent way will require more data and new methods. Although efforts should focus on improving countries’ capacity to collect and assess data, innovations in statistical methods and data collection technologies offer some potential solutions. Examples include multiple-imputation and small area estimation methods, technological innovations that enhance data collection, and statistical procedures for filling in data gaps. Multiple imputation can help ensure maximum survey sizes for poverty analysis by “filling in” missing data values with (a small number of) simulated alter- natives. Small area estimation techniques combine household survey and population census data to estimate poverty at more disaggregated levels and thereby pave the way for more precisely targeted policies. Chapter 5 provides details on these methods and emphasizes the importance of properly testing and validating such techniques. Complementary data for tracking poverty and shared prosperity across countries and over time Although household survey data are necessary for measurement of changes in poverty and shared prosperity, they are not sufficient. At a minimum, population data are also needed to convert survey-based estimates into national poverty counts and to make inferences about poverty for the population as a whole. Poverty assessments at the country level are usually denominated in local currency and based on a poverty line that is nation- ally determined. Cross-country comparisons therefore require additional data to count the poor across countries with a common currency and global poverty line. Purchasing power parity (PPP) indexes, produced by 20 OVERVIEW the International Comparison Program, perform this role. When survey data are not available on an annual basis, two additional sources of data are needed for comparisons of poverty across countries in a common (ref- erence) year: inflation data (to account for changes in prices between the survey year and the reference year) and real GDP growth data (to account for changes in real economic activity between the survey year and the refer- ence year). Data on prices are also needed to account for differences in the cost of living across different areas within countries. PPP indexes are a particularly important data input for cross-country comparisons of global poverty and shared prosperity. New rounds of PPPs can usefully update estimates of the cost of living across countries and provide price data for countries not previously covered. However, the introduction of new PPPs, which typically require reestimation of the international poverty line, can have substantial implications for the understanding of global poverty and can lead to significant reranking of countries and even regions. Table O.1 illustrates the sensitivity of the number of people who are estimated to be poor and the regional profile of poverty to changes in the PPP index and corresponding changes to the international poverty line. The 1993 count of the poor, based on the $1.08 poverty line and 1993 PPP numbers, estimated that 1.3 billion people were poor. Backcasting the 2005 PPP index to 1993, based on Table O.1 Estimates of the percentage poor in 1993, based on three PPP indexes 1985 1993 2005 Indicator or region ICP PPP index ICP PPP index ICP PPP index Poverty line $1.01 $1.08 $1.25 East Asia and the Pacific 26.0 25.2 50.8 Europe and Central Asia 3.5 3.5 4.3 Latin America and the Caribbean 23.5 15.3 10.1 Middle East and North Africa 4.1 9.0 4.1 South Asia 43.1 42.4 46.9 Sub-Saharan Africa 39.1 49.7 56.9 Poverty headcount 29.4 28.2 39.2 Poverty population (millions) 1,350 1,304 1,799 Source: Based on data from Deaton (2010). Note: ICP = International Comparison Program; PPP = purchasing power parity. 21 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY the new $1.25 poverty line, resulted in an estimated count of 1.8 billion people who were poor. Adopting the new index and revising the poverty line essentially resulted in increasing the estimated number of people who were poor in 1993 by 500 million. As discussed in chapter 6, careful review of new rounds of PPP indexes has been needed in the past. In some cases, new PPP indexes have not been adopted for global poverty measurement, while in other cases they have been adopted only after careful review and, at times, adjustments to correct for biases in the underlying data.7 In the case of the PPP rounds from 1993 and 2005, Chen and Ravallion (2001) and Ravallion, Chen, and Sangraula (2009) provide the evidence used to justify adopting these revisions to the PPP indexes. Given the substantial revisions in the development com- munity’s understanding of poverty across countries that can occur with a new round of PPP indexes, there is a need for caution and prudence in interpreting new PPPs indexes before they are applied to global poverty data. This kind of careful review of the recently released 2011 PPP indexes is currently under way and will need to be completed before a decision is made on whether and how to adopt them for global poverty estimation. Chapter 6 highlights the sensitivity of poverty estimates to the qual- ity of complementary data. For example, an estimated absolute error rate of 5 percent in population projections from census data could result in the poverty status of approximately 50 million people being misclassi- fied—leading not only to an error in the overall global poverty count, but potentially also to a distortion in the geographic profile of poverty across countries.8 Similarly, chapter 6 illustrates how the quality of inflation data can have profound consequences for the measurement of poverty in vari- ous countries. This insight not only reinforces the discussion above on the importance of high-quality input data for measuring poverty and shared prosperity, but also speaks to the importance of the entire data architecture at the country level. Given the importance of complementary data for pro- ducing poverty estimates, focusing on improved household surveys alone will not be enough. What are needed are well-developed national statistical systems that can collect robust, complementary data as well. Concerted effort is needed to improve measurement methods and data A well-functioning system of data sources and tools is needed to measure poverty and shared prosperity in a way that helps to monitor and improve 22 OVERVIEW policy. This report describes the detailed analysis needed to assess poverty and shared prosperity in a robust and consistent way. The report argues that although the World Bank’s twin goals are a useful and important umbrella around which essential distributional issues can be considered and discussed, the global perspective inherent in these goals may not necessarily coincide directly with the priorities of individual countries. This dual purpose of distributional analysis has been borne in mind in the discussion of this report. A key message of the report is that strengthening data measurement and collection capacity at the country level is of utmost importance. Although the World Bank’s goals are global, they will be achieved through policies at the national level, and the path to reaching the global poverty and shared prosperity goals will be heterogeneous across countries. The primary pur- pose of collecting data on extreme poverty and shared prosperity should therefore be to inform policy at the national level. The ability to make cross-country comparisons, while important, is secondary to having a solid evidence base to guide countries’ policies. This in turn implies that the data needs of national statistical agencies should not take a back seat to the data demands of international organizations, quality of data should not be compromised in favor of cross-country comparability, donors should accordingly be cautious in the emphasis they give to cross-country comparisons, and technologies and statistical approaches to bridge gaps in data measurement may offer some partial solutions. A useful comparison of different approaches to data collection can be made between the World Bank’s Living Standards Measurement Study (LSMS) and the United States Agency for International Development’s (USAID’s) Demographic and Health Surveys (DHS). The LSMS seeks to strengthen household data collected by national statistical agencies through intensive collaboration with those agencies, and focuses on improving data collection methods and developing an instrument that is tailored to the country context. One outcome of the focus on collection methods and capacity building has been that some countries have continued to manage and fund the survey even when external support has tapered off. A trade- off, however, of tailoring the instrument to the country context is that many of the indicators and data collected are frequently difficult to compare across countries. In contrast, the DHS data collection activities tend to be much more focused on defining and measuring indicators in a manner that maintains comparability across countries. Similarly, DHS training is rela- tively more focused on international standards for measuring an extensive set of key health and nutrition indicators. As a result of this, the DHS effort 23 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY has enjoyed great success in increasing the coverage and standardization of health data, which has promoted an explosion of detailed analysis of health issues in countries and has helped donors to make cross-country compari- sons and thereby target funding for health aid more effectively. To some extent, whether to emphasize cross-country comparability or the need to be sensitive to the country context is determined by the overarching purpose of the survey. In the case of DHS, the focus is on measures of health and well-being for which the items to be measured are similar across countries and units are the same (or conversion factors are well known). In contrast, a primary objective of the LSMS is to measure consumption poverty, which consists of items that vary significantly across countries (for example, rice in India, teff in Ethiopia) and units are not standardized internationally (for example, sacks and piles). The twin goals of shared prosperity and poverty reduction rest heavily on the measure of consumption, which does not lend itself as readily to standardization. Since the primary aim of measurement and collection of poverty and pros- perity data is to support policy making at the national level, strengthening the statistical systems of countries is a key priority. The LSMS approach of working closely with national statistical agencies has been important in this respect. Greater support to enhance the capacity of statistical agencies and more funding for improved data systems are needed. The timeliness and frequency of data collection need to increase. Even where data have been collected, processing lags can be lengthy and in some cases governments are reluctant to provide access. Even more emphasis needs to be placed on the importance of open access to data: it is regret- table that even in cases where data are produced some governments remain reluctant to make them available in an open and timely way. Beyond producing more frequent surveys, however, more attention to the careful design and collection of data is needed. There is ample scope to improve the standardization of data. Indeed, there needs to be more standardization of guidelines for estimating poverty and more emphasis on maintaining comparable measures of consumption and income. However, in many cases countries may have good reasons to follow a particular approach, which is different from that followed in other countries. Although this heterogeneity comes at the cost of comparability across countries, the benefits of data that can provide locally useful information may at times outweigh this cost. In all cases, the quality of national data, rather than its comparability, should be the primary concern. The implication is that donors and development practitioners need to be realistic about how much can be inferred from cross-country comparisons, not only because poverty and shared prosperity 24 OVERVIEW estimates may be imprecise, for many reasons discussed in the chapters that follow, but also because of heterogeneity in data across countries. This underscores the importance of informing funding decisions on the basis of a wide spectrum of evidence, rather than only a few indicators. New technologies and statistical approaches can help to bridge some of the gaps in data measurement and assessment. For example, technologi- cal innovations, such as computer-assisted personal interviews or mobile phone–based data collection, can help improve the frequency of surveys, especially in geographically dispersed countries. Similarly, when the desired or standard sources of complementary data are not available, it may still be possible to measure poverty with alternative data sources and mod- eling techniques. The use of technologies that can improve data collection and the use of well-designed survey-to-survey imputations should be scaled up. However, the first-best solution is to strengthen countries’ capacity to collect data in a manner that produces high-quality, time-sensitive, and well-documented inputs for policy making. The development community urgently needs to mobilize efforts to spur the availability of data for the purpose of poverty analysis (box O.5). Box O.5 Summary of the report’s key recommendations • As countries begin to consider what policy data are most relevant for policy. Strengthening changes will be needed to end poverty and the capacity of national statistical agencies boost shared prosperity, attention should be to collect these data should not be neglected given to the nature of growth in countries. in favor of data collection by international While strong and sustained growth will be criti- organizations. Data system architectures at cal to meeting the goals, attention to the type the country level are needed not only to support of growth (sustainable growth that benefits the credible measurement of the twin goals, but also poor) is also needed. for effective national development policy. • Rather than gauging progress toward the World • Quality of data should be the primary aim of Bank’s goals separately, progress toward the efforts to improve data measurement and col- two goals should be assessed in unison, as lection at the country level. Although increased “twin” goals. Achieving progress in both goals frequency of data is desirable, it should not come will require efforts on both fronts. at the cost of improved quality. Similarly, for the purpose of well-designed poverty mitigation • Measuring progress toward ending poverty and policies, producing high-quality data that suit boosting shared prosperity requires increased national contexts is more important than having capacity at the national level, where improved data that is comparable across countries. 25 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Notes 1. Chapter 1 sets out the World Bank’s approach to measuring global poverty in more detail. See Chen and Ravallion (2010) for a fuller description of how the $1.25 a day international poverty line was derived. 2. The scenarios explored in Ravallion (2013) reach qualitatively similar conclu- sions but are based on less restrictive assumptions. 3. See World Bank (1990) for a rich discussion of this issue. 4. The Green Revolution in India is a prominent example of an episode of exten- sive poverty reduction supported by growth that substantially improved the returns to agriculture. By contrast, high growth driven by commodity booms has not always translated to effective poverty reduction when the returns from extractive industries have remained concentrated in the hands of relatively few people. 5. The regional leader simulations in chapter 2 highlight some interesting differ- ences across regions. In particular, they suggest that changing the incidence of growth could have a substantial impact on poverty in Latin America and the Caribbean and South Asia, while the effect may be relatively smaller in Sub-Saharan Africa. This partly reflects that nature of the exercise: the distri- bution of income (consumption) of the regional leader in Sub-Saharan Africa (Rwanda) is significantly less progressive than that of the regional leader in Latin America and the Caribbean (Brazil), thus the simulated impact of the exercise for countries in Sub-Saharan Africa is smaller. These regional effects are muted in the second simulation where the income or consumption growth of the bottom 40 percent is raised by 1 and then 2 percentage points more than their country’s respective mean growth rates. See chapter 2 for a full discussion. 6. See World Bank (2012) for a comprehensive discussion of the importance of sustainable development for poverty reduction and future growth. 7. The first comprehensive attempt at producing global poverty estimates was completed in 1979, on the basis of the 1975 International Comparison Program (ICP) PPP data. Since then the 1985, 1993, and 2005 ICP PPP revi- sions have been incorporated into the World Bank’s global poverty estimates— although often with a delay of five or more years—but the 1980 revisions were not incorporated. See chapter 6 for a full discussion. 8. In an extensive review of population counts, the National Research Council (2000) assessed the overall quality of population projections across the globe. Across several sets of United Nations and World Bank forecasts, the absolute value of the errors in projected country populations averaged 4.8 percent in five-year projections and 17 percent in 30-year projections. References Beegle, Kathleen, Joachim De Weerdt, Jed Friedman, and John Gibson. 2012. “Methods of Household Consumption Measurement through Surveys: Experimental Results from Tanzania.” Journal of Development Economics 98 (1): 3–18. 26 OVERVIEW Birdsall, Nancy, Nora Lustig, and Christian J. Meyer. 2014. “The Strugglers: The New Poor in Latin America?” World Development 60 (August): 132–46. doi:10.1016/j.worlddev.2014.03.019. Chen, Shaohua, and Martin Ravallion. 2001. “How Did the World’s Poorest Fare in the 1990s?” Review of Income and Wealth 47 (3): 283–300. ———. 2010. “The Developing World Is Poorer Than We Thought, but No Less Successful in the Fight Against Poverty.” Quarterly Journal of Economics 125 (4): 1577–1625. Deaton, Angus. 2010. “Price Indexes, Inequality, and the Measurement of World Poverty.” American Economic Review 100 (1): 5–34. doi:10.1257/aer.100.5. Ferreira, Francisco H. G., Julian Messina, Jamele Rigolini, Luis-Felipe López- Calva, Maria Ana Lugo, and Renos Vakis. 2012. Economic Mobility and the Rise of the Latin American Middle Class. Washington, DC: World Bank. Jolliffe, Dean, Gaurav Datt, and Manohar Sharma. 2004. “Robust Poverty and Inequality Measurement in Egypt: Correcting for Spatial-Price Variation and Sample Design Effects.” Review of Development Economics 8 (4): 557–72. doi:10.1111/j.1467-9361.2004.00252.x. López-Calva, Luis F., and Eduardo Ortiz-Juarez. 2014. “A Vulnerability Approach to the Definition of the Middle Class.” Journal of Economic Inequality 12 (1): 23–47. Mahalanobis, Prasanta C. 1963. The Approach of Operational Research to Planning in India. Indian Statistical Series 18. New York: Asia Publishing House. National Research Council (United States). 2000. Beyond Six Billion: Forecasting the World’s Population. Edited by John Bongaarts and Rodolfo A. Bulatao. Washington, DC: National Academy Press. Ravallion, Martin. 2013. “How Long Will It Take to Lift One Billion People Out of Poverty?” The World Bank Research Observer 28 (2): 139–58. Ravallion, Martin, Shaohua Chen, and Prem Sangraula. 2009. “Dollar a Day Revisited.” The World Bank Economic Review 23 (2): 163–84. doi:10.1093/ wber/lhp007. World Bank. 1990. World Development Report 1990: Poverty. New York: Oxford University Press. ———. 2012. Inclusive Green Growth: The Pathway to Sustainable Development. Washington, DC: World Bank. 27 CHAPTER ONE Defining and Assessing the Goal of Ending Poverty by 2030 Monitoring poverty at the global level has been an important pillar of the World Bank’s analytical work on poverty since an early attempt in the late 1970s to estimate the fraction of the developing world’s population in pov- erty (Aluwahlia, Carter, and Chenery 1979) and a subsequent effort for the 1990 World Development Report (Ravallion, Datt, and van de Walle 1991). What has changed now is that the World Bank has set an explicit goal and timetable. In the spring of 2013, World Bank President Jim Yong Kim presented to the international development community the World Bank’s new goal of ending global poverty by 2030. Ending poverty was defined as occurring when global poverty has fallen to no more than 3 percent of the world’s population. Thus, the World Bank is focusing on reducing global poverty from an estimated 14.5 percent in 2011 to 3 percent or lower in a period of two decades.1 This chapter starts with a brief review of the World Bank’s approach to measuring global poverty. The World Bank’s methodology has been widely described and discussed and so the review here will be selective, focusing in particular on the data-intensive nature of the effort and the accompanying sensitivity of poverty estimates to changes in data and empirical methods. The chapter then turns to an assessment of what reaching the poverty goal will require, demonstrating that the goal of ending poverty by 2030 is highly aspirational. The chapter examines a range of scenarios and shows that “business as usual” at the country level is unlikely to bring the world all the way to the 3 percent target. This implies, therefore, that success will depend on transformational policies that succeed either in markedly raising growth rates in countries or in improving the responsiveness of poverty reduction to growth through greater inclusion of the poor in the growth process. 29 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY What do evidence and analysis tell us about the likely pathway that global poverty reduction will trace during the coming decades? Poverty has fallen at a fairly steady rate of about 1 percentage point per year between 1980 and 2011, with China having a central role in shaping this global picture. But maintaining such a pace of poverty reduction through growth alone will become increasingly difficult. The reason for this can be under- stood in a simple stylized framework that analyzes the impact of growth on poverty when changes in the distribution of income are ruled out. Of course, given that policy makers can intervene and income distribution can and does change, it is necessary to look beyond stylized examples to actual country experience. The chapter documents that, on the one hand, in many developing countries the poverty of certain subgroups of the population is relatively insensitive to overall rising income levels. These pockets of poverty can emerge for a variety of reasons. They can be linked, for example, to geo- graphic remoteness, patterns of social stratification and discrimination, as well as market failures that generate poverty traps. As overall poverty levels fall and these pockets come to represent the majority of those who remain poor, progress in further reducing poverty may slow. On the other hand, evidence of poverty decline among those countries that, today, have already ended poverty suggests that in some cases policy makers were able to adopt the policies needed to maintain a steady rate of progress in elimi- nating extreme poverty. It is clear that for the global target of 3 percent by 2030 to be achieved, countries will need to look beyond accelerating growth toward ensuring that the poor in particular benefit from growth. This message motivates the examination in chapter 2 of the World Bank’s second goal, which is to boost shared prosperity. A brief overview of global poverty measurement Measurement of poverty with household surveys The World Bank’s approach to measuring poverty has been widely docu- mented.2 Household surveys play a pivotal role in this effort; they are critical not only for global poverty estimation, but also for country-level poverty estimation. The surveys are organized at the country level and are commonly administered by government statistical agencies. The key indi- cator of interest from such surveys is a measure of household consumption or income. In what follows, the discussion provides a brief overview of the 30 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 procedure to put together a consumption measure. Additional remarks are provided on the distinction between income and consumption. In a typical survey, a nationally representative sample of households is interviewed and asked to specify purchases against a list of market prod- ucts over a given period of time. Information is also collected about con- sumption of nontransacted (home produced) goods and services, access to publically provided goods and services, as well as ownership of assets such as housing and consumer durables. By combining the responses to such questions, it is possible to arrive at an estimate of the level of consumption in each surveyed household.3 Sampling weights that accompany the house- hold survey data can be used to extrapolate from the sample data to the underlying population. The resultant data on the distribution of consump- tion across the population can be combined with a poverty line to identify the poor. Specific methodologies can be applied to aggregate up from the household-specific poverty indicators to a national-level poverty measure. Consumption per capita is the preferred welfare indicator for the World Bank’s analysis of global poverty. This position needs to be explained with respect to two important but distinct alternatives. First, it is well under- stood and widely acknowledged that poverty is a complex phenomenon that involves multiple dimensions of deprivation. So shouldn’t global pov- erty be measured in an explicitly multidimensional framework? A focus on consumption poverty could otherwise be construed to imply that the World Bank regards other, nonconsumption dimensions of deprivation as of secondary importance. Such an interpretation would be unfortunate. Recognition that there are multiple dimensions of deprivation does not mean that they are best assessed simultaneously, within a single indica- tor. There is a good deal of consensus that a comprehensive consumption aggregate captures many important economic dimensions of well-being. Other critical dimensions of well-being—such as health, education, social inclusion, empowerment, and so on—are difficult to incorporate into a consumption measure. This is because it is difficult to construct a multi- dimensional measure that respects households’ own perspectives on how the various dimensions interact with each other and the trade-offs between the dimensions. Stopping short of attempting to construct an explicitly multidimen- sional measure of poverty does not, however, mean that comparisons of deprivation along various dimensions are not possible. World Bank pov- erty assessments, carried out at the country level, routinely look not only at consumption poverty, but also at deprivation along other dimensions. 31 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY These studies draw important insights not only from quantitative house- hold survey data, but also from qualitative surveys and from techniques that “mix” qualitative and quantitative analysis. Only when all relevant indicators are closely scrutinized and considered can an overall assessment of poverty be regarded as complete. The key point is that the dimensions can be examined alongside one another and do not necessarily need to be combined in a single indicator—with the additional proviso that one must not lose sight of the possibility that individuals and households may suffer multiple deprivations simultaneously. At the global level, the challenges of estimating multidimensional poverty measures are further aggravated. However, the central point remains that global consumption poverty can and should be examined alongside deprivations in all other relevant dimen- sions. Thus, the Millennium Development Goals were articulated with respect to a wide range of indicators, in addition to global consumption poverty. Similarly, there are large literatures exploring the cross-country correlations of health and education outcomes with poverty outcomes, as well as a growing literature investigating the association between economic well-being and subjective assessments of welfare. Chapter 3 offers addi- tional discussion of some of the issues around multidimensional poverty measurement. The second alternative to a consumption-based analysis of poverty is to measure poverty on the basis of income. Income is a widely available alternative measure of economic well-being and can be calculated from many household surveys in a manner similar to the procedure followed for constructing a consumption measure. In some countries, notably in Latin America, income is the only available indicator of economic welfare. In these countries, the most common nationally representative household sur- veys are employment surveys, designed to collect information on employ- ment patterns and labor incomes. The surveys readily yield a measure of household income, but they provide no information on household con- sumption. In those instances where household income is the only possible indicator of economic well-being, the World Bank’s global poverty moni- toring effort uses income as the welfare indicator to measure a country’s poverty rate. However, income is not, in general, the preferred indicator. It can be argued that, in measuring poverty, policy makers are inter- ested in capturing the living standards achieved by individuals. These are directly reflected in a well-constructed, comprehensive consumption measure. Incomes, in contrast, reflect an opportunity to reach a given wel- fare but may provide only an imperfect proxy of what welfare level was 32 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 finally achieved. In the face of a particularly poor agricultural harvest, for example, a farmer might generate a very low or even negative income. But by drawing down on stocks and by borrowing from friends and relatives, it may be possible for the farmer’s family to maintain its consumption levels, at least for some time.4 Consumption may thus provide a smoother, less volatile measure of living standards than income, reflecting not only the financial inflows that are available to a household (as captured by a current income measure), but also the ability of a given household to (dis) save or borrow. More generally, the preference for consumption derives from the fact that these data are typically more easily and accurately collected in the developing country context. This is particularly the case when attention is focused on the poor. The poor are likely to consume a rather modest range of goods and services, primarily staple food items and a small set of essential nonfood goods and services. Compiling information on the consumption levels of the poor may thus be reasonably straightforward. By contrast, collecting information on the income levels of the poor can be much more complex. The poor are likely to be employed in the infor- mal sector and may derive income from multiple sources—each of which contributes in a small way to total income. If income is measured over a long period, like a year, it could be easy to overlook some of these income sources. Furthermore, many of the developing world’s poor are subsistence farmers with incomes that may be particularly difficult to calculate given long lags and uncertain attribution across seasons, between when costs of cultivation are incurred and when associated farming revenues accrue.5 Comparison of poverty across countries In the 1990 World Development Report on poverty, the World Bank applied a concerted effort to estimate the world’s population in poverty (World Bank 1990). As described in Ravallion, Datt, and van de Walle (1991), this effort was based on what was, at the time, a rather thin empirical foundation, consisting of a single household survey available in only 22 countries. The empirical base underpinning the World Bank’s global pov- erty estimates has since increased substantially. Chen and Ravallion (2010) estimated global poverty in 2005 based on 675 household surveys for 115 countries covering the period 1979 to 2006. Ravallion (2013), drawing on Chen and Ravallion (2013), estimated global poverty at three points in time between 1990 and 2008, drawing on 900 surveys for 125 countries. 33 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY The most recent World Bank estimates, covering the period 1981 to 2011, expand that database further, to well over 1,000 surveys covering nearly all developing countries.6 Assembly of the survey data that support the World Bank’s global poverty monitoring task is carried out by a designated team in the World Bank’s research department. The resultant internationally comparable poverty estimates are published in a database called Povcal. An accom- panying website and online computational tool called PovcalNet provide access to these data to users within and outside the World Bank.7 A key feature of PovcalNet is that users can access and manipulate the Povcal data remotely, either to replicate the World Bank’s calculations or to tailor the analysis to their own specific needs. The process governing the collection of household survey data in each country varies on a case-by-case basis. In some countries, household survey data collection is an integral part of the mandate of the national statistical office (NSO). Surveys are programmed into the NSO’s work plan, and survey-based consumption data are regularly published and disseminated. In many countries, however, there is no such systematic effort to collect and distribute survey data. Household surveys are collected on an ad hoc basis—as a result of specific requests from a particular government depart- ment or ministry and depending on the availability of funding. Often donors provide the impetus and funding for data collection. Even in those countries where survey data have been collected and compiled, there is great heterogeneity across countries as to when and to what degree the data are made available to analysts outside the NSOs. Delays between the fielding of household surveys and the release of the data for analysis can be lengthy. In quite a few countries, access to survey data remains altogether restricted. Occasionally, as in the case of China, even though access to microdata is restricted, aggregated data on the distribution of consump- tion is published in official NSO reports. In such cases, indirect estimates of poverty might still be feasible, although not without the imposition of additional assumptions. These considerations account for the lack of a consistent and predictable flow of new data into PovcalNet from all countries. Even in those cases where new data do become available, the fact that they stem from choices, decisions, and implementation at the country level implies that there are numerous ways in which comparability across countries of the underlying consumption data can be compromised. Before any calculation of global poverty can proceed, therefore, it is necessary to undertake an exhaustive assessment and evaluation of each country’s respective data. In some cases, 34 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 adjustments can be introduced so as to strengthen comparability; often, harmonization will be far from complete and a degree of noncomparability will remain. Some imprecision in the resulting global poverty estimates will be unavoidable. Scrutiny of the household surveys that enter the World Bank’s Povcal database occurs at multiple levels. First, household surveys are identified and acquired (and occasionally procured directly) by World Bank teams working in specific countries and regions. These data are checked and analyzed by the World Bank country teams for the purpose of national- level poverty work in the respective country and region. The survey data are then sent to the Povcal team and are subjected to a further round of scrutiny, this time from the point of view of their comparability with data from all other countries. Second, the Povcal team is in some cases able to identify and acquire household surveys that have been collected outside the purview of the World Bank’s country teams. Such data include surveys collected in devel- oped countries where there is no presence of a World Bank operational unit or data that have been collected by NSOs that do not have a dialogue with the World Bank’s operational units (for example, Iran). The way in which these additional data sources are accessed can range from a routine downloading of the data from officially approved websites (as is the case for many developed country data sets) to the acquisition of data via per- sonal networks and ad hoc requests. As these data have not undergone any World Bank scrutiny, they must be assessed by the Povcal team from first principles. The challenges with such data can be particularly onerous, as the quality of survey documentation received on an ad hoc basis may be quite variable. Third, several household surveys are collected by NSOs with substantial technical assistance from the World Bank research department’s Living Standards Measurement Study (LSMS). This program focuses on meth- ods of data collection with an eye toward providing guidance on the most appropriate methods for collecting data on living standards. The current LSMS-Integrated Surveys for Africa program involves the LSMS team in informing the collection of household survey panels in seven African coun- tries. More generally, some 80 surveys included in the research depart- ment’s global poverty monitoring effort have come directly from such data collection efforts involving the LSMS team. Although, again, there is no assurance that these data are strictly comparable across countries, the data have generally received close scrutiny by World Bank researchers and are usually well documented. 35 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Irrespective of how the data are sourced, they are all vetted and assessed for their suitability for cross-country comparability in the global poverty monitoring effort. Although considerable efforts are made, it is clear that comparability remains partial and will hopefully strengthen further over time, as additional methodological refinements are developed and introduced. Chapter 5 provides a further discussion of the challenges in assembling a global database of comparable country-level surveys. It describes some of the innovations under consideration aimed at increas- ing the frequency of country-level poverty estimates and the timeliness of global poverty estimates. National and global poverty lines Poverty analysis at the World Bank is most commonly carried out at the country level. It is important that this work is done in a way that is relevant to the respective country, producing empirical results that can be readily interpreted and endorsed by stakeholders, and providing reliable informa- tion to support decision making. Country-level analysis commonly builds on a solid measure of per capita consumption and combines this with a poverty line that has been derived in the respective country, representing a well-understood and widely accepted minimum threshold of consumption. The intention of such a poverty line is to delineate the threshold standard of living in a given society below which an individual is judged to be poor by the standards of that particular society. The standards are likely to vary across societies. Indeed, empirical evidence indicates that across countries, national poverty lines tend to rise with average income levels. Box 1.1 pro- vides a detailed overview of the various approaches that have been taken in the specification of national poverty lines. Global poverty estimates represent the sum of country-level estimates but are based on a common poverty line across all countries. The World Bank currently uses an international poverty line of $1.25 a day, in 2005 prices. As described in Chen and Ravallion (2010), this line corresponds to an average of the national poverty lines of the 15 poorest developing coun- tries and must therefore be understood to represent a very low threshold standard of living.8 In setting the global poverty line at $1.25 per person per day in real terms, the World Bank has elected to monitor global poverty by the standards that apply in the very poorest countries of the world. It is sobering that, even at that standard, there were about one billion poor people in the world in 2011. 36 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Box 1.1 Setting national poverty lines around the world Poverty lines are commonly used as cutoff points calorie requirement can be met with various food that delineate who in a country or region is con- baskets and, depending on the cost composition sidered poor at any given point in time, based on of the basket and local price levels, the resulting some predefi ned standard of living. The choice poverty line can vary widely (Pradhan and others of poverty line—what type and how it should be 2000; Haughton and Khandker 2009). set—depends on the local context and intended In addition to the food component (which gives use. In high-income countries, where absolute the so-called food poverty line), the overall poverty deprivation is less common, poverty lines are often line often also includes a nonfood component that relative—that is, they are defined in relation to the is added to reflect costs for housing, clothing, elec- overall distribution of income. For example, a pov- tricity, and so on. There are various ways to esti- erty line could be set as a percentage of the overall mate the nonfood component—and no consensus population mean or median income. In develop- on best practice. One way is to stipulate a second ing countries, where large parts of the population consumption bundle that reflects an adequate level cannot meet their basic needs, it often makes sense of nonfood items. Parallel to the approach for the to define some absolute standard and thus set an food component, that bundle could then be priced absolute poverty line. accordingly. In the absence of an objective caloric The challenge of defining an absolute poverty requirement, however, it is difficult to define “ade- line at the country level can be summarized by two quate” nonfood consumption needs. An alternative related questions. First, what is the adequate mini- approach to estimate the nonfood component is to mum level of well-being at which an individual is not divide the food component by the average share of considered poor in the specific local context (often food in total household expenditure (Orshansky called the referencing problem)? Second, how can 1963), although this approach raises the question the minimum amount of money that corresponds to of whether the food share of the average household, that level of well-being be identified (the identifi- a poor household, or a nonpoor household should cation problem)? Commonly, these two problems be used. are approached in what is called the cost of basic An alternative to the cost of basic needs approach needs method . This approach first stipulates a con- is the food energy intake method , which does not sumption bundle that is deemed adequate for basic require information on the prices of the goods that consumption needs in the local context and then are included in the estimated consumption bas- estimates the cost of this specific bundle. ket. Instead, this approach plots total household What is an adequate consumption bundle? One (food and nonfood) consumption expenditure or potential starting point is the average nutritional income against food consumption as measured in requirement for an individual to be in good health, calories per person per day to find the level at which often approximated to be 2,100 calories per person a household can meet its basic energy requirements. per day. Based on this food energy requirement, a However, this requires analysts to assume a rela- local consumption basket is compiled for a diet that tionship between household expenditure and food reflects the consumption habits of local households energy, and this approach does not lend itself to near the poverty line. The cost of this basket is esti- comparisons across time or regions. Yet another mated based on the prices of the various foodstuffs potential approach to set absolute lines is based that are included. This is not a trivial task, since the on asking people what minimum consumption or (continued) 37 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Box 1.1 Continued income level they need just to make ends meet. Empirically across countries, national absolute These subjective poverty lines remain relatively rare poverty lines tend to drift upward with average in practice, but they can be useful supplements to income, although for the very poorest countries the more objective measures. relationship is initially flat (Ravallion, Chen, and Conceptually, the cost of basic needs approach Sangraula 2009). The median poverty line across provides the most reliable framework to set national countries of Sub-Saharan Africa (using data from absolute poverty lines and is widely used in prac- around 2000) was roughly equal to the World tice. In a data set of national poverty lines compiled Bank’s international poverty line of $1.25 a day by the World Bank’s Global Practice for Poverty, (at 2005 purchasing power parity [PPP]). Across 38 of 45 national poverty lines set in low- and countries in Latin America and the Caribbean middle-income countries between 2001 and 2011 around 2010, the median national poverty line was were based on the cost of basic needs method. The a little over $4 per capita per day (at 2005 PPP). In Russian Federation is one of the few countries that contrast, in the United States in 2013, a household use the food energy method, while the remaining with two adults and two children under 18 years countries in Eastern Europe and Central Asia pre- old was considered poor if its daily income was dominantly rely on relative poverty lines. less than about $16 (at current 2013 prices, around The common practice in high-income countries $13.50 at 2005 prices). is to use relative lines. In the European Union, the Ultimately, the choice of a specific absolute or main poverty measure identifies as “at risk of pov- relative poverty line is a social and policy decision erty” all households that have net incomes of less that depends on the local context. No matter how than 60 percent of the national median. Similarly, precisely a specific poverty line is estimated, it is the Organisation for Economic Co-operation and important to keep in mind that living standards Development uses national median household of those just above the poverty line are not very income as a yardstick and applies thresholds of 50 different from those just below. In other words, percent and 60 percent. A noteworthy exception is nothing happens to individuals in terms of their the United States, where the federal poverty mea- consumption, income, health, or any other indica- sures are based on absolute thresholds. In 1963, tor when their income crosses an absolute poverty U.S. government statistician Mollie Orshansky line (Deaton 1997; Pritchett 2006). The key issue, calculated the cost of a minimum food diet and then, in setting an absolute poverty line is not its multiplied it by three to account for nonfood precise location, but to ensure comparability and expenditure. Since then, her results have been consistency across areas and over time. adjusted for inflation and today form the basis for Source: Based on Deaton (1997); Haughton and Khandker a detailed matrix of poverty lines, varying by family (2009); Ravallion (1988); and Ravallion, Chen, and Sangraula size, number of children, and so on. (2009). Global poverty counts The World Bank employs a specific measure of poverty in its calculations. It reports the extent of global poverty by calculating the percentage of the world’s population with a consumption or income level below the interna- tional poverty line. Producing global poverty counts in this way is intuitive 38 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 and easily communicated. Yet it has disadvantages as well. Notably, this manner of measuring poverty is insensitive to the fact that there may be great variation in living standards among the poor across countries. Two countries could record the same headcount rate of poverty, although in one country the poor have consumption levels far below the poverty line, while in the other the poor’s consumption levels are only just below the poverty line. Other poverty measures, more sensitive to differences in consumption levels among the poor, can be readily calculated and reported but are less easy to communicate. Chapter 3 provides further discussion of some of the alternative methods for measuring poverty. The three key steps involved in measuring global poverty can be sum- marized as follows: construct a survey-based measure of household con- sumption, define a global poverty line, and aggregate across households to calculate an overall measure of poverty. While these steps capture the overall process, there remain a few measurement details to consider fur- ther. First, although household surveys typically collect information on consumption at the household level, poverty headcounts seek to assess the poverty of individuals and the percentage of the population that is poor. Conventional practice is to divide household consumption by household size and attribute to each individual in the household a per capita con- sumption level accordingly. Those individuals whose per capita consump- tion level is below the poverty line (also expressed in per capita terms) are designated as poor. It is important to note that proceeding in this manner involves several important assumptions: that household resources are shared equally across family members; that family members have identical needs, such that two individuals with the same per capita consumption level enjoy the same living standard; and that there are no differential costs of reaching a given welfare level per person for households of different sizes. All three of these assumptions are unlikely to hold in practice. Yet it would be difficult to relax the assumptions in a way that is transparent and widely accepted. An imperfect but tractable solution is to maintain the assumptions but to subject all conclusions to sensitivity analysis in which the assumptions are in turn allowed to be relaxed. Country-level work along these lines indicates that conclusions as to, for example, the relative poverty of the elderly versus children, or of the particular vulnerability of widows, can be quite sensitive to these assumptions.9 Second, important adjustments must be made to the survey results to account for differences across places and time. Differences in the cost of living between countries must be accommodated by converting con- sumption levels in each country into comparable international prices, or 39 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY purchasing power parity (PPP) terms. In addition, consumption levels from a given survey may need adjustment if there are important differences in the cost of living across regions of a given country, or if the year of the survey does not coincide with the year for which global poverty is being estimated. In the interval between the two time periods, economic growth may have occurred, as well as changes in the cost of living. To line up the data for all countries to a given reference year, adjustments are introduced on the basis of national accounts data on consumption growth and con- sumer price indexes that capture the rate of inflation over time. Chapter 6 provides further details on the various price and growth adjustments. Third, to be certain that poverty estimates are based on accurate popu- lation figures, population census data are used to translate the poverty rates in a country into numbers of poor people based on population estimates. In many cases, the household survey data are accompanied by accurate population weights that allow for a direct conversion of survey-based counts to the underlying population. But often these weights are outdated or incomplete and must be adjusted with census data. Further discussion of these issues is provided in chapter 6. Assessment of the global poverty target Given the broad approach taken by the World Bank to measure global poverty, this section attempts to provide a perspective on the target to end poverty by 2030. As was noted above, the World Bank’s target is to end poverty by reducing global poverty to 3 percent or less. Why should a global poverty rate of 3 percent be interpreted to imply an ending of poverty? As discussed below, poverty in many countries remains extremely widespread. Reducing poverty to zero in such countries over any reason- able time frame would be extremely unrealistic. However, a global goal of zero poverty would require the elimination of poverty in each and every country. It is also important to acknowledge that at any moment in time, some churning is likely to be taking place in which some people, possibly for reasons beyond their control, fall into poverty, even if only temporarily. It is difficult to imagine a world in which nobody at all is poor. For these reasons, it seems reasonable to view global poverty as having effectively ended even if some frictional poverty remains at a very low level. Hence, the global target is 3 percent or lower. What is the current picture of poverty around the world? In 2011, it is estimated that about one billion people in the world had a consumption 40 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 level below the $1.25 a day global poverty line (table 1.1).10 This represents about 17 percent of the population of the developing world and 14.5 per- cent of the entire global population.11 In 2011, poverty was most prevalent in Sub-Saharan Africa and South Asia. These two regions accounted for about 80 percent of the global poor. Taking the 2011 poverty estimates as a point of departure, figure 1.1 illustrates in a stylized way the changing patterns of global poverty in selected developing regions. In each of the three panels, the vertical gray line indicates today’s $1.25 global poverty line, and the vertical axis can be read as the poverty headcount at each consumption or income level. In 1981 (panel a), the estimated number of people below the $1.25 line was about 1.9 billion, representing about 52 percent of the developing world’s popula- tion. Poverty was most prevalent in East Asia and the Pacific, where about 77 percent of the population had a consumption or income level below the $1.25 line, and South Asia, where about 61 percent of the population was considered poor. As the graph illustrates, the headcount in these regions was mostly driven by China and India; the total number of poor people in each of these two countries accounted for about three-quarters of the headcount in East Asia and the Pacific and South Asia, respectively. In Sub-Saharan Africa, slightly more than half the regional population was considered poor. Table 1.1 Poverty in 2011 at $1.25 a day 2005 PPP Headcount Number of poor Region (%) (millions) East Asia and the Pacific 7.9 160.8 Europe and Central Asia 0.5 2.3 Latin America and the Caribbean 4.6 27.6 Middle East and North Africaa 1.7 5.6 South Asia 24.5 399.0 Sub-Saharan Africa 46.8 415.4 Total developing world 17.0 1,010.7 World 14.5 1,010.7 Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: Benchmark year estimates generated following methodology described in Chen and Ravallion (2010). For countries without survey data, such as Eritrea or Somalia, the poverty headcount is assumed to be equal to the respective regional average headcount. a. This is a provisional estimate, in part because it is based on survey data covering only about one third of the population and because the 2011 estimate for Egypt is a projection from 2008 data. 41 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 1.1 Changing patterns of global poverty, 1981–2030 a. 1981 0.020 Density (share of population) 0.018 0.016 0.014 $1.25-a-day international poverty line 0.012 0.010 0.008 0.006 0.004 0.002 0 0 10 20 30 40 50 60 70 80 90 100 Consumption expenditure or income per capita per month (2005 PPP$) b. 2011 0.020 Density (share of population) 0.018 0.016 0.014 0.012 0.010 0.008 0.006 0.004 0.002 0 0 10 20 30 40 50 60 70 80 90 100 Consumption expenditure or income per capita per month (2005 PPP$) c. 2030 0.020 Density (share of population) 0.018 0.016 0.014 0.012 0.010 0.008 0.006 0.004 0.002 0 0 10 20 30 40 50 60 70 80 90 100 Consumption expenditure or income per capita per month (2005 PPP$) China India Sub-Saharan Africa East Asia and Pacific (excluding China) South Asia (excluding India) Latin America and Caribbean Source: Based on analysis of World Bank PovcalNet data. Note: Calculations assume lognormal distributions, based on the mean and variance of the country-specific consumption or income distributions in each reference year. To generate annual reference years, survey means are log-linearly interpolated between survey years and extended back- ward to 1980 and forward to 2011, using real household consumption growth. The graph for 2030 shows a projection under the assumption that countries grow between 2011 and 2030 at their historical average growth rates over the period 2000 to 2011 and constant inequality since the last available survey year. PPP = purchasing power parity. 42 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 In comparison, panel b in figure 1.1 illustrates the central role of India and China in the global reduction of poverty as measured by the $1.25 line over the following three decades. In East Asia and the Pacific, the headcount fell from about 77 percent of the regional population in 1981 to about 8 percent in 2011. In South Asia, the headcount more than halved, from 61 percent in 1981 to about 24 percent in 2011. Still, of the one billion people in the world below the poverty line in 2011, about 40 percent were in South Asia, about 30 percent in India alone. As shown in table 1.1, poverty in 2011 was most prevalent in South Asia and in Sub-Saharan Africa. In 2030 (figure 1.1, panel c), assuming historical, country-specific growth rates and no changes in the distribution of income, continued progress in global poverty reduction, particularly in South Asia, would likely leave Sub-Saharan Africa as the region with the highest poverty headcount. The rest of this section will now turn to a more detailed exami- nation of scenarios tracing out the possible evolution of poverty to 2030. In a comprehensive approach to projecting poverty declines, Ravallion (2013) poses the question of how long it would take to lift one billion people out of poverty. Projecting poverty rates forward on the basis of a variety of alternative growth scenarios, Ravallion found that achievement of this goal could take half a century or longer if a relatively pessimistic growth scenario is assumed. However, he found that this time span could be halved if the developing world could maintain the progress against extreme poverty that it was able to achieve during the first decade of the 2000s. Drawing on the line of reasoning outlined by Ravallion (2013), it is possible to construct a stylized scenario for reducing global poverty to 3 percent by 2030 by assuming, first, that each developing country grows at 4 percent per person per year (roughly equivalent to the average rate of growth of the developing world as a whole during the 2000s, as reflected in household survey data); second, that the distribution of consumption or income in each country remains unchanged throughout; and third, that between-country inequality remains unchanged. That means that popula- tion growth in each country is assumed to be equal to the global average population growth rate. Table 1.2 shows that, conditional on these assump- tions, poverty in the developing world will have fallen to 3.5 percent of the developing world’s population by 2030, corresponding to an overall global poverty rate of 3 percent.12 It is noteworthy that even with this growth scenario, poverty in Sub- Saharan Africa would remain at just over 19 percent in 2030, accounting 43 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Table 1.2 Ending global poverty Headcount Number of poor Region (%) (millions) East Asia and the Pacific 0.3 8.0 Europe and Central Asia 0.1 0.4 Latin America and the Caribbean 2.1 15.1 Middle East and North Africa 0.2 0.8 South Asia 1.3 24.9 Sub-Saharan Africa 19.2 202.5 Total developing world 3.5 251.8 World 3.0 251.8 Source: Based on analysis of World Bank PovcalNet data. Note: Values are for poverty in 2030 at $1.25 a day purchasing power parity (2005), assum- ing constant inequality and average per capita consumption growth of 4 percent per year. The projection assumes that each country’s survey mean per capita household income or consumption expenditure grows at about 4 percent per year, keeping between-country and within-country inequality constant. To keep between-country inequality constant, population growth rates in each country are assumed to be equal to the global average population growth rate between 2011 and 2030. This in effect shifts the entire world income or consumption expenditure distribution from 2011 to 2030. for nearly 80 percent of the global poor in that year. Poverty in South Asia, by contrast, would have fallen sharply, from about 24 percent in 2011 to about 1.3 percent in 2030. It is also of interest that, even with this assumed progress in poverty reduction over the coming two decades, sev- eral countries would remain with poverty rates above 30 percent. Scrutiny of country-specific poverty outcomes following the projections of this sce- nario shows that six countries would have poverty rates above 30 percent in 2030: Burundi (39 percent), the Democratic Republic of Congo (57 percent), Haiti (33 percent), Madagascar (52 percent), Malawi (37 percent), and Zambia (48 percent). As emphasized in Ravallion (2013), the assumption of a global per capita growth rate of 4 percent per year cannot be taken for granted. While the developing world as a whole achieved such progress during the past 10 to 15 years, many countries certainly did not grow at this rate. To gauge the realism of achieving the 3 percent target, it is thus instructive to consider a few additional growth scenarios. The discussion refrains, for the time being, from considering alternatives to the basic assumption that within-country inequality does not change. 44 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Ending global poverty: An ambitious target A first alternative simulation assumes that individual countries grow at their respective average annualized rates of the past 20 years, instead of a 4 percent growth rate in all countries. First, growth rates as captured in the national accounts are considered, before turning to discussion of growth rates that are based instead on household survey data. As is discussed in chapter 6, there is often less than perfect agreement between survey and national accounts data on aggregate consumption or income measures, as well as on growth rates, so it is useful to consider both sets of economic performance data in turn. Not surprisingly, table 1.3 reveals that when countries grow at their respective average national accounts growth rates of the past two decades, the poverty target appears more difficult to achieve. With this assumed growth performance, the global poverty estimate falls to 6.8 percent of the world’s population in 2030, much higher than the 3 percent achieved Table 1.3 Alternative One: Projections based on countries’ experiences over the past 20 years Headcount Number of poor Region (%) (millions) East Asia and the Pacific 0.4 8.5 Europe and Central Asia 0.1 0.7 Latin America and the Caribbean 3.1 22.1 Middle East and North Africa 1.6 7.0 South Asia 3.5 69.2 Sub-Saharan Africa 32.8 465.4 Total developing world 7.9 572.8 World 6.8 572.8 Source: Based on data from the World Bank PovcalNet database. Note: Values are for poverty in 2030 at $1.25 a day purchasing power parity (2005), assuming country-specific national accounts–based growth rates over the past 20 years. This projection assumes that each country’s mean per capita household income or consumption expenditure grows at past country-specific national accounts growth rates, keeping country- specific distributions constant. Past national accounts growth rates are calculated as the annualized growth rates of real gross domestic product per capita (countries in Sub-Saharan Africa) or the annualized growth rates of household final consumption expenditure per capita (all other countries) over the period 1990–2010. Endnote 14 provides a more detailed discussion. National accounts growth and population projections are based on the World Bank’s World Development Indicators database. 45 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY under the benchmark scenario. Under this scenario, the number of coun- tries with projected poverty rates above 30 percent in 2030 increases dra- matically from 6 to 23.13 Recognizing that (national accounts) growth picked up in most of the developing world from around 1999 onward, a third scenario retains the assumption of country-specific growth rates, but applies the annualized rate achieved during the past 10 years, rather than the past 20 years (table 1.4).14 This scenario yields a projected global poverty rate of 4.8 percent of the world’s population, a rate lower than in the preceding scenario, but still well above the aspirational global target of 3 percent. Under this scenario, too, some 17 countries would remain with projected poverty rates above 30 percent in 2030.15 Maps 1.1 and 1.2 illustrate the poverty rates in all developing countries in 2011 compared with the projected poverty rates in 2030 according to this scenario. The maps highlight the geographical concentration of poverty in Sub-Saharan Africa. As noted above, national accounts and survey-based annual growth rates for a given country can vary significantly (chapter 6 of this report provides Table 1.4 Alternative Two: Projections based on countries’ experiences over the past 10 years Headcount Number of poor Region (%) (millions) East Asia and the Pacific 0.3 5.6 Europe and Central Asia 0.0 0.1 Latin America and the Caribbean 3.2 22.6 Middle East and North Africa 1.1 5.0 South Asia 1.6 32.7 Sub-Saharan Africa 23.9 339.4 Total developing world 5.6 405.4 World 4.8 405.4 Source: Based on data from the World Bank PovcalNet database. Note: Values are for poverty in 2030 at $1.25 a day purchasing power parity (2005), assuming country-specific national accounts–based growth rates over the past 10 years. This projection assumes that each country’s mean per capita household income or consumption expenditure grows at past country-specific national accounts growth rates, keeping country- specific distributions constant. Past national accounts growth rates are calculated as the annualized growth rates of real gross domestic product per capita (countries in Sub-Saharan Africa) or the annualized growth rates of household final consumption expenditure per capita (all other countries) over the period 2000–10. Endnote 14 provides a more detailed discussion. National accounts growth and population projections are based on the World Bank’s World Development Indicators database. 46 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Map 1.1 Poverty headcount at $1.25 a day, 2011 50% and greater 40%—50% Estonia Russian Federation Latvia Russian Fed. Lithuania 30%—40% Poland Belarus Ukraine Kazakhstan Moldova Mongolia 20%—30% Georgia Uzbekistan Azer. Turkmenistan Kyrgyz Rep. D.P.R. of Armenia Korea 10%—20% Turkey Syrian Tajikistan Tunisia Lebanon China A.R. I.R. of Afghanistan 0%—10% Morocco West Bank and Gaza JordanIraq Iran Pakistan Algeria Nepal Bhutan Western Libya A.R. of 0% Sahara Egypt India Bang. Mexico Jamaica Cuba Myanmar Lao Cabo Verde Mauritania P.D.R. High-income economies Belize Guatemala Honduras Haiti Senegal Mali Niger Chad Sudan Eritrea Rep. of Yemen Thailand Vietnam The Gambia Burkina El Salvador Nicaragua Faso Nigeria Djibouti Cambodia Philippines No data Guinea-Bissau Guinea Benin Fed. States of Micronesia Marshall Is. Sri Ghana Costa Rica R.B. de Guyana Sierra Leone Togo Central South Ethiopia Lanka Panama Venezuela Suriname Liberia Côte Afr.Rep. Sudan Somalia Malaysia Palau Colombia French Guiana (Fr) d’Ivoire Cameroon Maldives Uganda São Tomé and Príncipe Gabon Congo Kenya Nauru Kiribati Ecuador Rwanda Kiribati Seychelles D.R.of Congo Burundi Papua New Solomon Tanzania Comoros Indonesia Guinea Islands Tuvalu Samoa Peru Brazil Timor-Leste Angola Malawi Zambia Fiji American Bolivia Vanuatu Samoa (US) Zimbabwe Madagascar Mauritius Fiji Tonga Paraguay Namibia Botswana Mozambique Dominican Swaziland Poland South Lesotho Republic Czech Rep. Africa Chile Uruguay Slovak Rep. Ukr. Argentina Hungary Dominica Slovenia Croatia Romania St. Vincent and St. Lucia Bosnia and the Grenadines Herzegovina Serbia Kos. Bul. Grenada Montenegro FYR Trinidad Albania Mac. R.B. de Venezuela and Tobago Antarctica IBRD 41110 Source: Based on World Bank PovcalNet data. Note: See notes to table 1.1 for further explanations. Map 1.2 Poverty headcount at $1.25 a day, 2030 50% and greater 40%—50% Estonia Russian Federation Latvia Russian Fed. Lithuania 30%—40% Poland Belarus Ukraine Kazakhstan 20%—30% Moldova Mongolia Georgia Uzbekistan Kyrgyz Rep. D.P.R. of Armenia Azer. Turkmenistan 10%—20% Turkey Syrian Tajikistan Korea Tunisia Lebanon China Morocco A.R. Iraq I.R. of Afghanistan 0%—10% West Bank and Gaza Jordan Iran Pakistan Algeria Nepal Bhutan Western Libya A.R. of 0% Jamaica Sahara Egypt India Bang. Mexico Cuba Myanmar Lao Cabo Verde Mauritania P.D.R. High-income economies Belize Haiti Mali Niger Sudan Eritrea Rep. of Yemen Thailand Vietnam Guatemala Honduras Senegal Burkina Chad Nicaragua The Gambia Cambodia Philippines El Salvador Faso Nigeria Djibouti Fed. States of Micronesia No data Guinea-Bissau Guinea Benin Sri Marshall Is. Ghana Costa Rica R.B. de Guyana Sierra Leone Togo Central South Ethiopia Lanka Panama Venezuela Suriname Liberia Côte Afr.Rep. Sudan Somalia Malaysia Palau Colombia French Guiana (Fr) d’Ivoire Cameroon Maldives Uganda São Tomé and Príncipe Gabon Congo Kenya Nauru Kiribati Ecuador Rwanda Kiribati Seychelles D.R.of Congo Burundi Papua New Solomon Tanzania Comoros Indonesia Guinea Islands Tuvalu Samoa Peru Brazil Timor-Leste Angola Malawi Zambia Fiji American Bolivia Vanuatu Samoa (US) Zimbabwe Madagascar Mauritius Fiji Tonga Paraguay Namibia Botswana Mozambique Dominican Swaziland Poland South Lesotho Republic Czech Rep. Africa Chile Uruguay Slovak Rep. Ukr. Argentina Hungary Dominica Slovenia Croatia Romania St. Vincent and St. Lucia Bosnia and the Grenadines Herzegovina Serbia Kos. Bul. Grenada Montenegro FYR Trinidad Albania Mac. R.B. de Venezuela and Tobago Antarctica IBRD 41110 Source: Based on World Bank PovcalNet data. Note: Projections of poverty rates are based on countries’ experience over the past 10 years. See table 1.4 for further details on projection methodology. 47 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY a further discussion). India represents a striking illustration of such diver- gences. As described in several entries in Deaton and Kozel (2005), a large discrepancy exists between National Accounts Statistics (NAS) estimates of household consumption expenditure in India and estimates derived from the National Sample Survey (NSS), with the survey-based estimates pointing to much lower consumption levels. Of particular concern is that this gap has been growing over time, with NSS estimates of consumption growth significantly lower than NAS-based estimates. In light of these considerations, an additional scenario gauges the plau- sibility of a 3 percent global poverty target in 2030 by abstracting from national accounts growth. Instead this scenario considers average growth rates over the past 10 years calculated directly from the survey data. In a few cases, there are gaps in survey data availability or problems with comparability across surveys. In those cases, growth rates remain based on national accounts estimates. Table 1.5 illustrates that when historical survey-based growth rates are applied, global poverty declines to just under 6.7 percent by the year 2030. This is significantly higher than the 4.8 percent that would be achieved if Table 1.5 Alternative Three: What do household surveys say? Headcount Number of poor Region (%) (millions) East Asia and the Pacific 1.0 21.7 Europe and Central Asia 0.1 0.7 Latin America and the Caribbean 2.9 20.3 Middle East and North Africa 0.9 4.0 South Asia 2.4 47.5 Sub-Saharan Africa 33.2 470.7 Total developing world 7.8 564.8 World 6.7 564.8 Source: Based on data from the World Bank PovcalNet database. Note: Values are for poverty in 2030 at $1.25 a day purchasing power parity (2005), assuming country-specific household survey–based growth rates over the past 10 years. This projection assumes that each country’s mean per capita household income or consumption expenditure grows at past country-specific household survey growth rates, keeping country- specific distributions constant. Past survey growth is calculated over a period of about 2000 to 2012. When survey growth is not available, national accounts growth rates are used as described in endnote 14. In 11 countries where survey growth rates over that period are negative, a survey mean growth rate of 1 percent per year is assumed. National accounts growth and population projections are based on the World Bank’s World Development Indicators database. 48 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 national accounts growth rates of the past 10 years had been applied and accords with expectations that survey-based estimates of aggregate con- sumption, and of consumption growth, are often (but not always) lower than national accounts estimates. A final scenario explores a more aspirational setup, again using historical national accounts growth, but aiming to preserve a degree of plausibility. This scenario examines episodes of growth (covering periods of 8 to 10 years) during the past 20 years, in each country in turn. The scenario identifies the growth rate associated with the episode of most rapid growth during this reference period and postulates that the country will manage to match that growth rate going forward over the coming two decades. Although this is the basic principle that underpins this scenario, several filters are applied to ensure that the scenario remains broadly plausible.16 Table 1.6 reveals that with these ambitious, but not entirely unachiev- able, growth projections, the target of 3 percent poverty in the world comes tantalizingly close to being within reach. This growth scenario is predicated Table 1.6 Alternative Four: An aspirational scenario Headcount Number of poor Region (%) (millions) East Asia and the Pacific 0.1 0.9 Europe and Central Asia 0.0 0.1 Latin America and the Caribbean 2.9 20.7 Middle East and North Africa 0.4 1.8 South Asia 0.6 12.0 Sub-Saharan Africa 21.0 297.4 Total developing world 4.6 332.9 World 4.0 332.9 Source: Based on data from the World Bank PovcalNet database. Note: Values are for poverty in 2030 at $1.25 a day purchasing power parity (2005), assum- ing country-specific household survey–based growth rates associated with the highest 10-year growth episode observed during the past 20 years. This projection assumes that each country’s mean per capita household income or consumption expenditure grows at the rate achieved during the best past country-specific national accounts growth spell, keeping country-specific inequality constant. Past national accounts growth spells are calculated as the annualized growth rates of real gross domestic product per capita (countries in Sub-Saharan Africa) or the annualized growth rates of household final consumption expenditure per capita (all other countries) over 8 to 10 years, observed during 1992–2012. When the best annual growth rate is less than 1 percent per year, a growth rate of 1 percent per year is assumed. Endnotes 14 and 16 provide a more detailed discussion. National accounts growth and population projections are based on the World Bank’s World Development Indicators database. 49 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY on growth rates that would need to be sustained over the next two decades but that do, at least, have some historical precedence in each respective country. If all countries were to manage to grow at these rates, then this simulation exercise indicates that a global poverty rate of 3 percent is close to being achievable. If growth can be accelerated further or inequality can be brought down, then the goal becomes more readily attainable (see chap- ter 2). At the same time, there should be no mistaking that a requirement of sustained growth at the very high levels postulated by this scenario, over a period of two decades, is quite onerous and is one that historical experience suggests is far from certain (chapter 4 explores this point further). Country-specific projections It is instructive to consider how countries will need to adjust, on a country- by-country basis, to come close to the projected 3 percent global poverty estimate in 2030. Table 1.7 and figure 1.2, panel a, report for the 10 coun- tries currently contributing the most to global poverty in 2011 the rates Table 1.7 Actual and required growth rates in the 10 countries contributing most to poverty in 2011 Current Number of poor Current growth Required growth Country headcount (%) (millions) rate (%) rate (%) India (Rural) 25.5 213.9 3.52 3.52 India (Urban) 22.9 87.5 3.94 3.94 Nigeria 60.1 98.6 2.28 3.17 China (Rural) 12.3 81.7 7.73 7.73 Bangladesh 39.6 60.5 2.32 4.97 Congo, Dem. Rep. 84.0 53.7 1.51 1.65 Indonesia 16.2 39.5 4.29 6.43 Ethiopia 36.8 32.9 1.72 3.42 Pakistan 12.4 21.8 3.68 5.52 Tanzania 43.5 20.2 1.35 5.75 Philippines 19.3 18.3 1.43 7.92 Source: Based on data from the World Bank PovcalNet database. Note: Values reflect those needed to achieve the aspirational scenario presented in table 1.6, which assumes country- specific household survey–based growth rates associated with the highest 10-year growth episode observed during the past 20 years. 50 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Figure 1.2 What does it take? Actual and required growth rates to achieve the aspirational scenario a. Countries contributing most to poverty in 2011 b. Countries contributing most to poverty in 2030 Congo, India, rural 3.5 3.5 Dem. Rep. 1.5 1.7 India, urban 3.9 3.9 Nigeria 2.3 3.2 Nigeria 2.3 3.2 Madagascar 0.8 1.0 China, rural 7.7 7.7 Kenya 1.6 2.0 Bangladesh 2.3 5.0 Malawi 3.4 4.6 Congo, Dem. Rep. 1.5 1.7 Zambia 2.9 3.3 Indonesia 4.3 6.4 Burundi 0.5 2.2 Ethiopia 1.7 3.4 Niger 0.9 2.0 Pakistan 3.7 5.5 Côte d’Ivoire 0.0 1.0 Tanzania 1.4 5.8 Philippines 1.4 7.9 India, rural 3.5 3.5 Number of 0 2 4 6 8 Number of 0 1 2 3 4 5 6 poor people poor people Historical and required growth Historical and required growth in 2011 in 2030 (percent) (percent) (million) (million) 200 million 60 million 20 million 10 million Historical Required Source: Based on data from the World Bank PovcalNet database. of growth that would be needed in each country respectively to comply with the “aspirational” scenario discussed above. The scenario requires that in Nigeria, currently contributing heavily to global poverty numbers, growth rates need to rise, but not to a degree that is entirely unimaginable. In Nigeria, the growth rate needs to pick up from 2.3 to 3.2 percent. More onerous challenges are confronted by countries such as Tanzania (having to raise growth from a current 1.4 percent to 5.8 percent) and the Philippines (requiring an increase from 1.4 to 7.9 percent per year). Table 1.8 and figure 1.2, panel b, report the growth rates that are required for the 10 countries that, in the year 2030, will remain as the main contributors to global poverty. If all developing countries man- age to achieve their historically highest growth rates, then in 2030 the Democratic Republic of Congo will be the single largest contributor to global poverty, accounting for 64.3 million poor people. To comply with 51 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Table 1.8 Actual and required growth rates in the 10 countries that will remain as principal contributors to global poverty in 2030 2030 Number of poor Current growth Required growth Country headcount (%) (millions) rate (%) rate (%) Congo, Dem. Rep. 62.0 64.3 1.51 1.65 Nigeria 22.5 61.5 2.28 3.17 Madagascar 73.3 26.4 0.82 1.01 Kenya 23.2 15.4 1.62 2.02 Malawi 56.5 14.7 3.37 4.59 Zambia 51.9 13.0 2.88 3.34 Burundi 75.4 12.4 0.53 2.21 Niger 27.8 9.6 0.89 2.0 Côte d’Ivoire 29.8 8.7 0 1.0 India 0.6 8.1 3.52 3.52 Source: Based on data from the World Bank PovcalNet database. Note: Values reflect those needed to achieve the aspirational scenario presented in table 1.6, which assumes country- specific household survey–based growth rates associated with the highest 10-year growth episode observed during the past 20 years. the aspirational scenario, the Democratic Republic of Congo will need to have lifted its annual growth rate, as captured in household survey data, from 1.5 to 1.7 percent as captured in household survey data. Nigeria will contribute another 61.5 million poor people to the global total. To achieve the aspirational scenario, the country will need to have raised its annual growth rate from 2.3 to 3.2 percent. Does the ending poverty target become more elusive when nearing success? One of the important features of the past three decades of global poverty reduction has been that poverty has been declining at a steady rate of approximately 1 percentage point per year. Figure 1.3 shows that there has been a striking linearity in the decline of the global headcount index since the early 1980s. In his original exploration of the prospects for 52 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Figure 1.3 Poverty reduction in the developing world, global measures 1980–2010 60 Poverty measures at $1.25 a day 50 Poverty 40 headcount (percent) 30 Poverty 20 gap 10 0 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2010 Sources: Based on data from the World Bank PovcalNet database and Chen and Ravallion (2013). ending global poverty, Ravallion (2013) assumed that the developing world would achieve and sustain a growth rate in per capita consumption of just over 4 percent per year. Coupled with an additional assumption of uniform population growth in all countries, Ravallion’s (2013) analysis implied that global inequality was also assumed not to change over the decades to 2030. If one were to apply the assumptions in Ravallion (2013) to, say, the world’s income distribution in 1990 and then project poverty forward to 2010, a similar path of global poverty reduction would be traced as was empirically observed during these two decades. How realis- tic is it to assume that, going forward from 2011, poverty reduction will maintain this constant pace? Why might the pace of poverty reduction be lower in the future? Suppose that the world could be thought of as just one large country and that per capita consumption in this country was growing at a steady rate, without any accompanying change in distribution. Why would one expect eventually to see a declining rate of poverty reduction in this setting? Empirically, distributions of consumption take a shape that reflects a con- centration of the population around the middle of the income distribution, with a thinning of the population density around either of the two tails. 53 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY As was described earlier, one can think of a poverty line as representing a fi xed standard of living indicated by a particular consumption level. Persons with consumption levels below that line are considered poor, while those whose consumption is higher are counted among the nonpoor. As the distribution of consumption shifts along the consumption scale (the result of economic growth that essentially scales up the consumption levels of all persons in the population, which is needed to satisfy the assumption of unchanged inequality), it is clear that poverty will fall as economic growth progresses. However, because the bulk of the population is concentrated around the middle of the consumption distribution, it is also the case that, progressively, the fraction of the population that are lifted out of poverty as a result of economic growth will decline. To put it simply, a large number of people tend to live on consumption levels near the average, while rela- tively fewer live on very high or very low consumption levels. After poverty reduction has reached the mass of people concentrated in the middle of the consumption distribution, poverty reduction will increasingly reach fewer people, even if the pace of growth remains unchanged. Figure 1.4 plots this empirical result for two stylized distributions of per capita consumption. For illustrative purposes, consumption is transformed to a normal distribution by taking the logarithm of consumption.17 The two panels show how many people hold what level of consumption: panel a plots the share of individuals at each point of the distribution while panel b plots the cumulative population share. At the point where the curve in panel b intersects the poverty line, the vertical axis provides the total poverty headcount. The earlier period is represented by the red curve t 0 while the later period is represented by the orange curve t 1. Between the two periods, all consumption levels increase by the same proportion, so that inequality remains constant. In other words, each person gets richer by the same rate (but richer individuals gain more in absolute terms than poorer individuals). As a result, the curve of log consumption shifts to the right but the shape remains the same. Consider the earlier period t 0. In figure 1.4, panel a, a large share of the population is lifted out of poverty as the peak of the red curve moves beyond the poverty line. In panel b, the same shift of the red curve means that the poverty line is crossed at the point where the curve is very steep. Now consider the later period t 1, where growth remains the same but the majority of the population has already escaped poverty. In panel b, the curve crosses the poverty line at a relatively less steep point, which means that the same growth in consumption will translate into a less marked decline in the poverty headcount. The slope at 54 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Figure 1.4 The effect of growth on poverty under the assumption of unchanged inequality a. Kernel density estimate Poverty line Log income or consumption expenditure t0 t1 b. Cumulative percentage of population 100% Poverty line 80% 60% ⌬ Poverty reduction 40% ⌬ Income 20% 0 Log income or consumption expenditure t0 t1 Note: Panel a illustrates a stylized density of income, that is, the share of individuals at each point of the distribution, plotted on a logarithmic scale. Panel b illustrates the cumulative population share of the same distribution on the same logarithmic scale. The two angled lines in panel b represent the slope of the tangent at the point where the poverty line intersects with the cumulative distribution function, which is the point that marks the total poverty headcount for each distribution. each of the two intersections (dashed lines) can be thought of as the growth elasticity of poverty reduction.18 Thus, under the assumed conditions of unchanging inequality and constant growth, there exists virtually a mechanical relationship between growth and the sensitivity of poverty reduction to that growth. In the face of such conditions, the only way that a constant rate of poverty decline can 55 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY be delivered all the way to zero is if growth, in fact, accelerates over time. Yoshida, Uematsu, and Sobrado (2014) offer the striking finding that, if one were to assume that the world consumption distribution as existed in 2010 could be regarded as that of just one global country, then to maintain a steady rate of global poverty reduction all the way to zero by 2030, global growth would have to accelerate to rates as high as 48 percent per year by the end of the period (figure 1.5). The discussion in this section so far has assumed that, somehow, all countries would grow at the same rate. Yet it was emphasized in the pre- ceding section that a plausible assessment of the global poverty targets must allow for growth rates across countries to vary. Given that, histori- cally, growth rates have differed across countries, it was argued above that imposing a single, uniform growth rate across all countries is an important departure from realism. What do heterogeneous growth rates imply for the trajectory of poverty reduction over time? In the previous section it was shown that, if growth rates are allowed to vary across countries, only the aspirational scenario Figure 1.5 Declining sensitivity of poverty reduction would require ever-increasing growth 50 20 45 18 Poverty headcount at the end of period Annual growth rate required (percent) 40 16 35 14 30 12 (percent) 25 10 20 8 15 6 10 4 5 2 0 0 2 4 6 8 0 2 4 6 8 0 –1 –1 –1 –1 –2 –2 –2 –2 –2 –3 10 12 14 16 18 20 22 24 26 28 20 20 20 20 20 20 20 20 20 20 Required growth rate (left y-axis) Poverty headcount (right y-axis) Source: Based on data from the World Bank PovcalNet database. 56 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 would yield a global poverty rate by 2030 in the vicinity of the 3 percent target. This scenario assumed constant growth rates at the country level that were sufficiently high to ensure that the reduction in poverty achieved each year and in each country was on aggregate sufficient to bring global poverty down to very nearly 3 percent by 2030. The precise trajectory of poverty reduction that this scenario implies is displayed in figure 1.6. In the figure, progress toward the 3 percent target in 2030 occurs not as a linear reduction of poverty over time, but rather through a trajec- tory of a gradually declining pace of poverty reduction as overall global poverty falls. The reason for the diminishing pace of global poverty reduction observed in figure 1.6 can be understood in terms of the differential pace of growth, and hence poverty reduction, across countries. Some countries, such as China, with initially high levels of poverty, but also very rapid growth rates, would initially see rapid rates of poverty reduction over time. Initially, because China contributed substantially to the global poverty count, global poverty would fall commensurately with China’s falling poverty. Over time, however, continued growth and poverty reduction in China would translate into a progressively smaller impact on global Figure 1.6 The trajectory of future poverty reduction may not be obviously linear 18 Global poverty headcount (percent) 16 14 12 10 8 6 4 3% target 2 0 30 10 12 14 16 18 20 22 24 26 28 20 20 20 20 20 20 20 20 20 20 20 Distribution-neutral simulation under aspirational scenario Hypothetical linear trend Source: Based on data from the World Bank PovcalNet database. 57 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY poverty, because China’s poverty would account for an ever smaller con- tribution to global poverty. At some point prior to 2030, China can be expected to have essentially eliminated poverty at the global poverty line, so further growth in China would have no further impact on global pov- erty. Thus, as the faster-growing countries, such as China, essentially grow themselves out of any further contribution to global poverty reduction, it is clear that the pool of countries accounting for the remaining global poverty will, on average, be growing less rapidly and will be reducing poverty at a slower pace. Pockets of poverty and the dynamics of poverty reduction at the country level Just as it is difficult to imagine that global poverty would decline at a constant rate all the way through to 2030, one would not expect poverty decline at the country level to display a straight-line trajectory. Within countries, too, there are reasons why poverty decline might be expected to slow with continued economic growth. Poverty reduction is an uneven process. In many countries, pockets of poverty exist whereby certain parts of the population appear not to be participating to the same extent as oth- ers in the broader development process. Country-level poverty assessments, in which household survey data are broken down to assess the poverty status of various groups, routinely identify specific groups—defined by education, occupation, ethnicity, race, religion, region of residence, and so on—as experiencing higher than average odds of being counted among the poor. In some cases, there may be poverty traps—situations in which it is not possible for the poor to extricate themselves from their disadvan- taged condition because of failures in credit, land, or other key markets or because low levels of education, skill levels, or health prevent them from availing themselves of the new opportunities offered by a general expansion of economic activity. In some societies, systematic patterns of discrimination and exclusion can be observed, linked to such factors as gender, ethnicity, origin, caste, religion, or race. Climate change may also be a factor that contributes to increased pock- ets of poverty in the future. As is discussed in chapter 4, climate change can lead to structural changes in the economy, and the growth elasticity of poverty may change if sectors and natural resources that the poor rely more heavily on are severely affected by climate change. An increasing number of studies suggest that climate change will disproportionately 58 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 affect the poorest because of their dependence on natural resources and ecosystem-based services—as a means to support consumption and accu- mulate assets, and as a form of safety nets that strengthen resilience in the face of shocks—because the poorest live in regions where the impact of climate change is expected to be most severe, or because they have less capacity (financial, institutional, and technical) to adapt to climate change.19 More broadly, changing climatic patterns may affect agriculture, biodiversity, and access to water, among other things, in a very localized manner, thus disproportionately affecting some areas or groups of people. Green growth strategies that aim to decouple economic growth from the emission of greenhouse gases can play an important role in limiting the disproportional effect of climate change on the poor.20 Empirical evidence of the existence of poverty traps, formally under- stood as self-reinforcing mechanisms that prevent the poor from escaping poverty, is difficult to assemble and remains scarce. Kraay and McKenzie (2014) survey the literature on the existence of poverty traps at the level of countries as well as among population groups within countries. They find relatively little evidence for the truly stagnant incomes that would be predicted by canonical models of poverty traps. At the same time, they acknowledge that this should not be taken to imply that poverty cannot be persistent among certain population groups. In fact, Kraay and McKenzie (2014) point to plentiful evidence of pockets of poverty related to geo- graphic location, arising from people being trapped in low-productivity locations, such as remote rural regions or low-productivity countries. The mechanisms underpinning such geographic poverty traps can vary across settings, but are usually related to physical and geographic characteristics that prevent households’ consumption levels from rising over time. Jalan and Ravallion (2002) discuss, for example, how in China the productivity of farmers’ investments is lower in poor areas, constraining their ability to lift themselves out of poverty. The poverty of people in such poverty pockets, whether they arise from proper poverty traps (representing a low-level equilibrium) or simply result from low resource endowments or patterns of discrimination, does not necessarily decline hand-in-hand with economic growth. As a result, as overall economic development generates employment and lifts the bulk of the population out of poverty, a core subset of the population may remain poor and constrained in its ability to benefit from growth. When poverty is spatially concentrated, one might expect overall growth to benefit some locations more than others and for the spatial concentration of poverty 59 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY to become increasingly accentuated over time. This will then reduce the responsiveness of aggregate poverty to further growth and will translate into a declining rate of poverty reduction with respect to growth. This process can be readily illustrated with the evidence recently assem- bled for Vietnam by Lanjouw, Marra, and Nguyen (2013). Two poverty maps were constructed for Vietnam, providing a snapshot of poverty at the district level in 1999 and 2009. During this time period, aggregate poverty in Vietnam fell sharply, from around 47 percent to 15 percent. Although poverty declined overall, some districts grew much more slowly than other districts and saw a much lower rate of poverty reduction than elsewhere. In other words, the spatial concentration of poverty increased notice- ably (map 1.3). For example, districts in the Red River Delta region, surrounding the city of Hanoi, and in the south of the country saw significant reductions in poverty, while the mountainous regions in the northwest and along the central coast and highlands of Vietnam saw slower progress. The growing spatial concentration revealed by the two poverty maps for Vietnam suggests that lagging districts will experience relatively less economic progress over time and will fall progressively behind the leading Map 1.3 Increased spatial concentration of poverty in Vietnam, 1999 and 2009 a. 1999 b. 2009 Northern Northern mountains mountains Red Red River Delta River Delta Central coast Central coast Central Central Poverty rate (%) highlands highlands 0–10 10–20 20–30 30–40 Southeast Southeast 40–50 50–60 60–70 70–80 80–90 Mekong Mekong 90–100 Delta Delta Source: Lanjouw, Marra, and Nguyen (2013). 60 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 regions. Heterogeneity in growth rates in turn implies that aggregate poverty will fall more slowly for a given overall national growth rate than would be the case if all districts grew at the same rate. Figure 1.7 illustrates this point by postulating alternative spatial patterns of growth for Vietnam over the decade from 2009 to 2019. In the first simulation, per capita expenditures are projected to grow at the same national rate of growth as was observed over the interval between 1999 and 2009. Poverty falls from just below 20 percent in 2009 to essentially zero by 2019. In subsequent simulations, household expenditure levels within a given region, province, or district are assumed to grow at the respective regional, provincial, or district-level growth rate that was observed between 1999 and 2009. Aggregate poverty reduction slows increasingly as the level of spatial het- erogeneity is allowed to increase. In sum, just as at the global level countries vary in their growth rates and their pace of poverty reduction, the existence of factors that result in unevenness in the rate of poverty reduction across population groups within a country, whether they are defined in terms of location of resi- dence or some other criterion, can result in a declining responsiveness of poverty reduction to a given rate of aggregate growth. The presence of such factors at the country level will thereby also translate into a declining Figure 1.7 Heterogeneous subnational growth in Vietnam leads to slower national poverty reduction 20 18 16 Growth rate (percent) 14 12 10 8 6 4 2 0 2008 2010 2012 2014 2016 2018 2020 National growth Regional growth Provincial growth District growth Source: Based on data from Lanjouw, Marra, and Nguyen (2013). 61 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY responsiveness of global poverty reduction to growth and will serve to delay achievement of the global poverty target. It is important to note that so far the discussion has ignored the pos- sibility of changes in the distribution of income. However, there is, of course, no cast-iron law stating that the distribution of income will remain unchanged in the future. Indeed, as will be shown in chapter 2, there have been important changes in the distribution of income in many countries— with inequality increasing in some and decreasing in others. Once one allows for changes in growth rates and inequality, the trajectory of poverty decline may be different from the scenarios described above. Changing underlying economic forces and policy choices may also substantially influence the future trajectory of poverty reduction. Some economic forces will play out on the global stage, but the relationship between growth and poverty decline will also often be driven by factors occurring at the country level or even at the subnational level. What does past country experience suggest about the likely pace of poverty reduction in the future? Although the discussion so far suggests that there are multiple reasons to expect poverty decline to slow with continued growth, the extent to which this may occur can vary. In principle, arguments in favor of an accelerating rate of poverty decline can also be made. As countries grow richer, they may move away from (imperfect) targeting of social policies and transfers toward universal entitlement programs that reach previously excluded pop- ulations. Alternatively, over time countries may acquire stronger admin- istrative data systems that can better implement means-tested programs. It is instructive to ask whether, empirically, it has been the case that in those countries where, today, poverty (at the global poverty line) has been eliminated, the pathway followed was inevitably one of poverty ending with a soft landing. In other words, does the experience of these countries support the idea that the pace of poverty reduction tends to slow over time? Or were there also cases where poverty ended more abruptly? Figure 1.8 depicts progress in poverty reduction over the very long run (Ravallion 2014, forthcoming). World poverty has fallen from an esti- mate of well above 80 percent in the beginning of the 19th century to under 20 percent today. The trajectory of poverty reduction in countries like the United States and the United Kingdom appears to have followed a fairly steady rate of decline until rates in the vicinity of 5 percent were 62 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Figure 1.8 Poverty reduction in countries that have already achieved zero extreme poverty, 1820–2000 80 Poverty headcount (percent) 60 40 20 0 1800 1850 1900 1950 2000 World United States United Kingdom Japan Austria, Czechoslovakia, Hungary and Ireland Italy Australia, Canada, New Zealand Germany Source: Based on data from Ravallion (2014) and Ravallion (forthcoming). Note: Based on estimates using parameterized Lorenz curves calibrated to the data set developed by Bourguignon and Morrisson (2002). See Ravallion 2014 for a more detailed explanation achieved. Subsequent poverty declines appear to have occurred more slowly. However, the path of poverty decline achieved in Japan—and in Austria, Czechoslovakia, and Hungary—appears to have followed a steady decline to zero. Although poverty ended later in these countries than in the United States and the United Kingdom, the percentage point decline in poverty was maintained to the end, and there is less clear evidence of a tapering off of poverty reduction in these countries. A more current picture of progress from the developing world can be observed in the case of Thailand. Figure 1.9 shows that in Thailand poverty estimated at the $1.25 global poverty line fell to 3 percent around 1995. Poverty reduction in Thailand prior to 1995 occurred at a constant rate and, indeed, there is some sign of acceleration relative to progress prior to 1990. As in the cases of Japan and Austria-Czechoslovakia-Hungary, there is little sign of a declining rate of progress in approaching the 3 percent poverty rate. Interestingly, after 1995, further progress in reduc- ing poverty in Thailand displays the familiar tapering-off tendency, with evidence of a small increase in poverty around 2000, at the time of the 63 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 1.9 Poverty reduction in Thailand, 1981–2010 45 40 Poverty headcount (percent) 35 30 25 20 15 10 5 3% target 0 1980 1985 1990 1995 2000 2005 2010 Poverty line at $1.25 a day Poverty line at $1.5 a day Poverty line at $1.75 a day Poverty line at $2 a day Source: Based on data from the World Bank PovcalNet database. East Asian crisis. Note further that if an alternative poverty line of $1.50 or $1.75 is taken, poverty in Thailand can be seen to end (reach 3 percent) only in the second half of the 2000s. The trajectory of the poverty line in these cases is less steep than was observed in the early 1990s. However, again, there is no evidence that the pace of poverty reduction diminished in the years immediately prior to the dates when poverty reached the 3 percent target. A detailed diagnosis of how the Thais were able to ensure that progress in poverty reduction was maintained all the way to the effective elimina- tion of poverty is beyond the scope of the present discussion. The point to emphasize, however, is that the actual experience of poverty reduction achieved by countries can be affected by policies and policy choices that bear not only on growth, but also on the distribution of growth. In other words, the rate at which poverty falls can be influenced by countries them- selves. This chapter argues that, most likely, achievement of the 2030 goal of ending global poverty will depend on countries pursuing a combination of growth and distribution policies that aim to supplement the attenuat- ing impetus that can be expected from continued growth alone. Efforts to boost the incomes of the lower deciles in particular are needed. This discussion is taken up further in chapter 2. 64 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Poverty and shared prosperity Reducing world poverty has long been an overarching objective of the World Bank. In recent years, the institution has taken the additional step of setting an explicit goal for itself and for the development community. A target has been articulated: to reduce global poverty to no more than 3 percent of the world’s population by 2030. This chapter has examined this target with a view toward gauging just what it means and what it implies. In particular, the chapter has attempted to assess whether business as usual, a continuation of recent global trends in poverty reduction, can be expected to be sufficient to reach the global poverty target. In thinking about the progress necessary for ending global poverty, it is important to have a clear understanding of the yardstick against which progress will be measured. This chapter has revisited the basic steps and procedures involved in the measurement of global poverty. A key concern has been to emphasize that not only are there many conceptual issues and choices involved in establishing a tractable means to monitor global pov- erty, but the quality and reliability of the underlying information database is critical. The development community continues to face important chal- lenges in strengthening its ability to gauge the extent of global poverty and how it is evolving over time. Several of these challenges will be examined in further detail in subsequent chapters of this report. This chapter has indicated that the explorations of poverty trends and trajectories that fed into the initial definition of the 3 percent global target were based on stylized assumptions—not necessarily intended to represent reality. The chapter has attempted to assess whether a path to the 3 percent target can be readily discerned once somewhat more realistic assumptions are made. In particular, the analysis in this chapter has focused on the kinds of growth scenarios that might unfold over the coming decades. Notably, the chapter has argued that these scenarios must accommodate different rates of growth across countries. The analysis cautions against complacency with regards to the achievability of future growth rates that are somehow anchored to historical experience and that are also able to generate poverty trajectories leading to an end of poverty by 2030. The chapter thus argues that the 3 percent target should be viewed as ambitious and far from assured. The chapter discusses further the point that, although past experience suggests that global poverty reduction up to around 2011 was achieved at a fairly constant pace of around 1 percentage point per year, there are many reasons to doubt that such a steady rate of poverty decline is feasible 65 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY as global poverty approaches the end target of 3 percent or lower. In the absence of targeted policies to the contrary, the pace of poverty decline associated with a given rate of economic growth can be expected, at some point, to diminish markedly. If this occurs well before the global target is reached, the burden on growth as an engine of global poverty reduction will become very significant: ever-increasing growth rates would be needed to maintain the overall pace of global poverty reduction. An important assumption maintained throughout the analysis in this chapter has been that income distribution within countries does not change. This assumption was needed to establish a benchmark against which the actual experience of countries can be gauged. Although the arguments pointing to a declining sensitivity of poverty reduction to growth follow readily when income distribution is held fixed, things are less clear once inequality is allowed to vary. Clearly, if inequality increases alongside economic growth, progress in poverty reduction may slow fur- ther. This chapter has illustrated this possibility in the context of uneven within-country spatial patterns of poverty reduction, where the persistence of pockets of poverty in a country results in a slower overall rate of poverty reduction. However, if inequality were somehow to fall alongside economic growth, particularly inequality associated with the relative position of the poor in the income distribution, then the rate of progress in poverty decline might hold steady or even accelerate as overall poverty approached zero.21 Indeed, there does seem to be some evidence of this from the experience of present-day countries in which extreme poverty has been eliminated. How this was achieved is likely to depend on a constellation of highly context-specific circumstances and policies. What is clear, though, is that if the incomes of the bottom segments can be boosted alongside overall economic growth, the prospects for ending global poverty are much enhanced. This observation motivates the focus in the next chapter on the World Bank’s second goal of boosting shared prosperity. Notes 1. Several additional details about the World Bank’s poverty goal are worth noting. First, the reference population for the poverty estimates is that of the world as a whole. Thus, poverty is to be brought down to under 3 percent of the world’s population, not that of the developing world only. Second, although 3 percent by 2030 is the ultimate objective, there is a need to galvanize energies and efforts right away. Accordingly, at the World Bank and International Monetary Fund Annual Meetings held in the autumn of 66 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 2013 in Washington, DC, the World Bank pointed to an interim target of 9 percent by 2020. 2. Ravallion, Datt, and van de Walle (1991); Chen and Ravallion (2004, 2010); Ravallion, Chen, and Sangraula (2009); and Ravallion (2013) describe the World Bank’s methods in considerable detail. Broader surveys of pov- erty measurement can be found in Ravallion (1994), Deaton (1997), and Ravallion (forthcoming). A useful overview of the World Bank’s approach can be found in Chandy (2013). 3. Deaton and Zaidi (2002) provide a useful guideline for constructing a com- prehensive measure of consumption. 4. Of course, difficult-to-interpret situations can also arise with consumption. For example, Lanjouw and Stern (1991) discuss the case of a rich individual in a small Indian village who chose to live (and consume) like an ascetic for religious reasons and who would have been categorized as poor on the basis of a consumption measure. 5. Note that in assessing the relative appeal of consumption over income as an indicator of economic well-being, one should not lose sight of the fact that when the focus moves from an assessment of welfare to an analysis of driv- ers of change, then income data become extremely useful as they can point to the differential roles played by different income sources (wages, remit- tances, profits). This underscores that deliberation around strategies for data collection may well wish to look beyond the gathering of reliable data on consumption only. 6. Prominent exceptions include the Democratic People’s Republic of Korea, as well as fragile and conflict-affected states such as Eritrea, Somalia, and Somaliland. 7. For details, see http://iresearch.worldbank.org/PovcalNet. 8. The 15 countries are mainly in Sub-Saharan Africa and comprise Chad, Ethiopia, The Gambia, Ghana, Guinea-Bissau, Malawi, Mali, Mozambique, Nepal, Niger, Rwanda, Sierra Leone, Tajikistan, Tanzania, and Uganda. See Chen and Ravallion (2010). 9. See Lanjouw, Lanjouw, Milanovic, and Paternostro (2004); Lanjouw and Ravallion (1995); and Drèze and Srinivasan (1997). 10. All regional and global poverty estimates in this chapter are based on an internal working version of the World Bank’s PovcalNet database with data as of August 2014. While every effort was made to use the most up- to-date data available, the estimates presented in this report should be seen as approximate; the official global and regional poverty estimates will be published in the forthcoming Global Monitoring Report. Concurrent with the publication of the Global Monitoring Report, the World Bank’s PovcalNet website will be updated. 11. The estimation of the global poverty headcount assumes that nobody lives below the $1.25 a day in high income countries. Although there are a number of people with household incomes below $1.25 per person in rich countries, estimated per capita consumption is above this threshold for nearly every- one. For example, Chandy and Smith (2014) find that 1 to 4 percent of the population in the United States live below $2 a day when using income 67 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY measures, but that fewer than 0.1 percent live below this threshold when using consumption data. 12. The procedure outlined in Ravallion (2013) does not insist on between- country inequality remaining constant, and therefore that all countries should grow at the same rate. We impose the additional assumption here in order to contrast it with the additional scenarios that follow. 13. These countries are Benin, Burundi, the Central African Republic, Comoros, the Democratic Republic of Congo, Republic of Congo, Côte d’Ivoire, The Gambia, Guinea, Guinea-Bissau, Haiti, Kenya, Madagascar, Malawi, Mali, Niger, Nigeria, Rwanda, Sierra Leone, Swaziland, São Tomé and Principe, Togo, and Zambia. 14. For the projections summarized in tables 1.3, 1.4, 1.5, and 1.6, annualized growth rates from national accounts are used to adjust mean per capita house- hold income or consumption expenditure from household surveys, keeping country-specific distributions constant. Following Chen and Ravallion (2010), national accounts growth rates are based on household final consumption expenditure per capita for all countries outside Sub-Saharan Africa. In Sub- Saharan Africa, real gross domestic product per capita growth is used because of better data availability. Before applying national accounts growth rates to survey mean per capita income and consumption expenditure, an adjustment factor of 0.87 times the growth rate is used to reflect empirically observed differences between national accounts growth and survey growth. Lower fac- tors are used for China and India (0.72 and 0.57, respectively) to account for historically larger gaps between national accounts and survey growth there. These factors are based on a simple cross-country regression for the growth rate in survey means on the growth rate from national accounts. Using an adjustment factor to account for such differences is common practice in related exercises. Birdsall, Lustig, and Meyer (2014) and Chandy, Ledlie, and Penciakova (2013) make a similar adjustment, but use different scaling factors. See Ravallion (2003) and Deaton (2005) for a more detailed discussion of the differences between national accounts growth and survey growth. 15. These are Benin, Burundi, the Central African Republic, Comoros, the Democratic Republic of Congo, Côte d’Ivoire, The Gambia, Guinea, Guinea-Bissau, Haiti, Liberia, Madagascar, Malawi, Mali, Swaziland, Togo, and Zambia. 16. The fi lters are as follows: (a) If the observed growth rate during the high- growth episode exceeded 10 percent per year, this was regarded as a case of noncomparable survey data and the case was dropped in favor of the second highest growth episode during the reference period. (b) In three cases, national accounts data revealed an even higher growth rate than that from the surveys; in these three cases, the high-growth rate, higher than 10 percent, was used. (c) If scrutiny of growth rates suggested that the surveys were noncomparable, they were replaced with national accounts estimates. (d) When the highest growth rate was positive but below 1 percent per year, a high growth of 1 percent was applied. (e) In the case of rural India, a high- growth rate equal to the rate achieved in the high-growth episode observed for urban India was applied. (f ) If growth rates over the 10-year window 68 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 were always negative, a positive growth rate from an episode shorter than the 10-year high-growth episode was considered. If there was no positive growth observed anywhere, a high-growth rate of 1 percent was assigned. 17. Empirically, per capita income often takes a shape in which there is a con- centration of the population around the middle of the distribution with a thinning around the two tails. In most cases, the mass of the distribution is concentrated on the left side (right-skewed). Such distributions can be approximated with a lognormal distribution, which means that the logarithm of income will be approximately normally distributed. Lopez and Servén (2014) show empirically that lognormal distributions provide a very close approximation of actual per capita income distributions. 18. See Bourguignon (2003) for a more formal treatment of the growth elasticity of poverty reduction. 19. For example, a recent study by Angelsen and others (2014) measures environ- mental incomes, showing that environmental income shares are higher for low-income households than the rest. 20. See World Bank (2012) for a comprehensive discussion of green growth strategies. 21. One potential indication of this would be if economic growth was greater for the poor relative to the country average growth rate. Chapter 2 introduces the shared prosperity indicator, which measures growth for the bottom 40 percent of each country. In those countries for which we have data, in about two thirds of the countries, this growth rate is greater than the overall aver- age growth rate. References Ahluwalia, Montek S., Nicholas G. Carter, and Hollis B. Chenery. 1979. “Growth and Poverty in Developing Countries.” Journal of Development Economics 6 (3): 299–341. doi:10.1016/0304-3878(79)90020-8. Angelsen, Arild, Sven Wunder, Ronnie Babigumira, Brian Belcher, Jan Börner, and Carsten Smith-Hall. 2014. “Environmental Incomes and Rural Livelihoods: A Global-Comparative Assessment.” World Development. In press. Birdsall, Nancy, Nora Lustig, and Christian J. Meyer. 2014. “The Strugglers: The New Poor in Latin America?” World Development 60 (August): 132–46. doi:10.1016/j.worlddev.2014.03.019. Bourguignon, François. 2003. “The Growth Elasticity of Poverty Reduction: Explaining Heterogeneity across Countries and Time Periods.” In Inequality and Growth: Theory and Policy Implications, edited by Theo S. Eicher and Stephen J. Turnovsky, 3–26. CESifo Seminar Series 1. Cambridge, MA: MIT Press. Bourguignon, François, and Christian Morrisson. 2002. “Inequality among World Citizens: 1820–1992.” American Economic Review 92 (4): 727–44. doi:10.1257/00028280260344443. 69 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Chandy, Laurence. 2013. “Counting the Poor: Methods, Problems, and Solutions Behind the $1.25 a Day Global Poverty Estimates.” Investments to End Poverty Working Paper 3, Development Initiatives, Bristol, U.K. Chandy, Laurence, Natasha Ledlie, and Veronika Penciakova. 2013. “The Final Countdown: Prospects for Ending Extreme Poverty by 2030.” Global Views Policy Paper 2013-04, The Brookings Institution, Washington, DC. Chen, Shaohua, and Martin Ravallion. 2004. “How Have the World’s Poorest Fared since the Early 1980s?” The World Bank Research Observer 19 (2): 141–69. ———. 2010. “The Developing World Is Poorer Than We Thought, but No Less Successful in the Fight Against Poverty.” The Quarterly Journal of Economics 125 (4): 1577–1625. doi:10.1162/qjec.2010.125.4.1577. ———. 2013. “More Relatively-Poor People in a Less Absolutely-Poor World.” Review of Income and Wealth 59 (1): 1–28. Deaton, Angus. 1997. The Analysis of Household Surveys: A Microeconometric Approach to Development Policy . Baltimore, MD: The Johns Hopkins University Press. ———. 2005. “Measuring Poverty in a Growing World (or Measuring Growth in a Poor World).” Review of Economics and Statistics 87 (1): 1–19. doi:10.1162/0034653053327612. Deaton, Angus, and Valerie Kozel. 2005. “Data and Dogma: The Great Indian Poverty Debate.” The World Bank Research Observer 20 (2): 177–99. doi:10.1093/wbro/lki009. Deaton, Angus, and Salman Zaidi. 2002. “Guidelines for Constructing Con- sumption Aggregates for Welfare Analysis.” Living Standards Measurement Study (LSMS) Working Paper 135, World Bank, Washington, DC. Drèze, Jean, and P. V. Srinivasan. 1997. “Widowhood and Poverty in Rural India: Some Inferences from Household Survey Data.” Journal of Development Economics 54 (2): 217–34. Haughton, Jonathan Henry, and Shahidur R. Khandker. 2009. Handbook on Poverty and Inequality. Washington, DC: World Bank. Jalan, Jyotsna, and Martin Ravallion. 2002. “Geographic Poverty Traps? A Micro Model of Consumption Growth in Rural China.” Journal of Applied Econometrics 17 (4): 329–46. doi:10.1002/jae.645. Kraay, Aart, and David McKenzie. 2014. “Do Poverty Traps Exist?” Policy Research Working Paper 6835, World Bank, Washington, DC. Lanjouw, Jean O., Peter Lanjouw, Branko Milanovic, and Stefano Paternostro. 2004. “Relative Price Shifts, Economies of Scale and Poverty During Economic Transition.” Economics of Transition 12 (3): 509–36. Lanjouw, Peter F., Marleen Marra, and Cuong Nguyen. 2013. “Vietnam’s Evolving Poverty Map: Patterns and Implications for Policy.” Policy Research Working Paper 6355, World Bank, Washington, DC. Lanjouw, Peter, and Martin Ravallion. 1995. “Poverty and Household Size.” Economic Journal 105 (433): 1415–34. Lanjouw, Peter, and Nicholas Stern. 1991. “Poverty in Palanpur.” The World Bank Economic Review 5 (1): 23–55. 70 DEFINING AND ASSESSING THE GOAL OF ENDING POVERTY BY 2030 Lopez, Humberto, and Luis Servén. 2014. “A Normal Relationship? Poverty, Growth, and Inequality.” Annals of Economics and Finance 15 (2): 593–624. Orshansky, Mollie. 1963. “Children of the Poor.” Social Security Bulletin 26 (7): 3–13. Pradhan, Menno, Asep Suryahadi, Sudarno Sumarto, and Lant Pritchett. 2000. “Measurements of Poverty in Indonesia—1996, 1999, and Beyond.” Policy Research Working Paper 2438, World Bank, Washington, DC. Pritchett, Lant. 2006. “Who Is Not Poor? Dreaming of a World Truly Free of Poverty.” The World Bank Research Observer 21 (1): 1–23. doi:10.1093/wbro/ lkj002. Ravallion, Martin. 1988. “Expected Poverty Under Risk-Induced Welfare Variability.” The Economic Journal 98 (393): 1171–82. doi:10.2307/2233725. ———. 1994. Poverty Comparisons . Fundamentals of Pure and Applied Economics 56. Chur, Switzerland: Harwood Academic Publishers. ———. 2003. “Measuring Aggregate Welfare in Developing Countries: How Well Do National Accounts and Surveys Agree?” The Review of Economics and Statistics 85 (3): 645–52. ———. 2013. “How Long Will It Take to Lift One Billion People Out of Poverty?” The World Bank Research Observer 28 (2): 139–58. doi:10.1093/ wbro/lkt003. ———. 2014. “Poverty in the Rich World When It Was Not Nearly So Rich,” Blog Post, Center for Global Development, Washington, DC. http://international .cgdev.org/blog/poverty-richworld-when-it-was-not-nearly-so-rich. ———. Forthcoming. The Economics of Poverty: History, Measurement, Policy. New York and Oxford: Oxford University Press. Ravallion, Martin, Shaohua Chen, and Prem Sangraula. 2009. “Dollar a Day Revisited.” The World Bank Economic Review 23 (2): 163–84. doi:10.1093/ wber/lhp007. Ravallion, Martin, Gaurav Datt, and Dominique van de Walle. 1991. “Quantify- ing Absolute Poverty in the Developing World.” Review of Income and Wealth 37 (4): 345–61. doi:10.1111/j.1475-4991.1991.tb00378.x. World Bank. 1990. World Development Report 1990: Poverty. New York: Oxford University Press. ———. 2012. Inclusive Green Growth: The Pathway to Sustainable Development. Washington, DC: World Bank. http://elibrary.worldbank.org/doi/book /10.1596/978-0-8213-9551-6. Yoshida, Nobuo, Hiroki Uematsu, and Carlos E. Sobrado. 2014. “Is Extreme Poverty Going to End? An Analytical Framework to Evaluate Progress in Ending Extreme Poverty.” Policy Research Working Paper 6740, World Bank, Washington, DC. 71 CHAPTER TWO Understanding Shared Prosperity Economic development is often equated with average growth in gross domestic product (GDP) per capita. A country with a high growth rate is deemed to be successful, while a country with a low or negative growth rate is considered to be falling behind. But this concept of development provides little insight into who may be benefiting (or not) from growth in a given country. For example, incomes in two countries could be growing at the same average rate, but in the first country all the growth is concen- trated among the richest members of society, while in the second, most of the growth benefits the poorest. Although both countries are progressing at the same pace, not everyone in each country benefits to the same extent. How should economic development be measured and assessed then? Should the emphasis be only on the overall performance of a country or should there be some additional focus to ensure that the poor are not left behind? This question, in essence, is what the World Bank’s new goal of boosting shared prosperity aims to address. The idea is to retain an empha- sis on growth (as measured in national surveys by income or consumption, as opposed to growth measured from national accounts), but shift attention to the growth of the average income (or consumption) of the bottom 40 per- cent of the people in a given country. In this way, shared prosperity remains an indicator of economic dynamism and progress and the benchmark of good progress is income growth of the bottom 40 percent of society. By introducing this goal, the World Bank explicitly brings income inequality to the forefront of the policy dialogue. The shared prosperity goal is not an inequality measure in itself, since it focuses exclusively on income growth of the bottom 40 percent of the population. However, as this chapter discusses, shared prosperity is intimately linked to inequality. The evolution of shared prosperity can be decomposed into a part that can 73 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY be attributed to overall (survey mean) income (or consumption) growth, while another part accounts for changes in the share of overall income that accrues to the bottom 40 percent. This simple relationship implies that shared prosperity reveals a lot about overall income distribution and, therefore, inequality. Although distributional issues have been part of the development debate for decades, this is the first time that an indicator with close links to inequality of outcomes (incomes) has become a benchmark of development progress for the World Bank. This chapter discusses the conceptual and empirical underpinnings of the World Bank’s shared prosperity goal. The chapter highlights some of the historical antecedents leading to the adoption of the shared prosperity goal and discusses the empirical requirements and challenges for calculat- ing shared prosperity trends in countries around the world. The chapter also presents some of the insights and benefits from tracking the income growth rate of the bottom 40 percent and argues why this measure is rel- evant even in countries where poverty is low. Finally, the chapter shows how the goal of shared prosperity goes hand in hand with the goal of ending global poverty and how boosting shared prosperity may be instru- mental for achieving the poverty goal. The chapter also discusses some of the data challenges in measuring and monitoring shared prosperity (which are discussed in more detail in chapters 5 and 6). The evolution of shared prosperity The desire to establish a measure that captures the notion that economic growth should benefit everyone is not new by any means. The debate on why and how to define such a measure can be traced to the 1950s. The discussion within the developing community on the concept of shared prosperity reached an important milestone in the early 1970s, as part of the process of refining the broad goals of development (box 2.1). One impor- tant influence on development thinking came from John Rawls’ Theory of Justice, with its implication that, in a society, promoting the well-being of the least fortunate member should be an important priority (Rawls 1971). An influential book published by the World Bank, Redistribution with Growth, by Chenery and others (1974), set in motion an active academic and policy debate on how best to measure such a concept. The authors argue that overall economic growth (measured by growth in gross national product [GNP]) is too narrow and cannot adequately be used as a social 74 UNDERSTANDING SHARED PROSPERITY Box 2.1 The World Bank’s early discussions of shared prosperity The development community’s rich debate on be done without unacceptable penalties to inclusive growth in the 1970s is reflected in a the concomitant goal of national growth. speech given by then World Bank President Robert Without specific emphasis on such pro- McNamara during the Annual Meetings of the grams, there will not be significant progress Board of Governors in 1972. in reducing poverty within acceptable time In the speech, McNamara first motivates the periods. importance of focusing on the bottom 40 percent Finally, he introduces the idea of monitoring of the population: “. . . the poorest 40 percent of and benchmarking: the citizenry is [a population] of immense urgency since their condition is in fact far worse than . . . The first step should be to establish national averages suggest.” specific targets, within the development He then calls for action: plans of individual countries, for income growth among the poorest 40 percent of the . . . Policies specifically designed to reduce population. I suggest that our goal should be the deprivation among the poorest 40 per- to increase the income of the poorest sections cent in developing countries are prescriptions of society in the short run—in five years—at not only of principle but of prudence. Social least as fast as the national average. In the justice is not merely a moral imperative. It longer run—ten years—the goal should be is a political imperative as well . . . it is pos- to increase this growth significantly faster sible to design policies with the explicit goal than the national average. of improving the conditions of life of the poorest 40 percent of the populations in the developing countries and that this can Source: Based on McNamara (1972). welfare indicator. By showing how growth in GNP can be decomposed into the growth of the incomes of socioeconomic groups with weights pro- portional to the groups’ existing share in the national product, the book presents a policy dilemma: in the process of maximizing overall growth, the best strategy would be to focus on those groups whose original share of GNP was the largest (in other words, the richest). As an alternative, the authors propose looking at the income growth performance of the poor to address the concern of maximizing social well-being. In this way, the concept of the growth performance of the bottom 40 percent and the goal of inclusive growth were introduced, although not widely used, as concrete measures emerging from this work. By the 1980s, a new strand of the literature emerged that contributed to the broadening of the goals of development and, as a consequence, the measurement of these goals. Led by the writings of Amartya Sen and 75 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY consistent with Rawlsian concepts, these pieces argue that access to or ownership of material goods should not be the goal of development (Sen 1983, 1985, 1999). Instead, development and progress should be seen in terms of functioning (what a person manages to do) and capability (what a person is able to do). Sen points out that this approach goes back to the work of Adam Smith and Karl Marx, but it was lost in the increasing effort to measure the progress of nations by their incomes. A direct consequence of this work was the emergence of a broad range of nonmonetary measures of living standards, such as the United Nations Development Programme’s Human Development Index or, more recently, the Oxford Poverty and Human Development Initiative’s multidimensional poverty index and the World Bank’s Human Opportunity Index (see chapter 3 for a full discus- sion on these and other alternative welfare measures). This perspective was also implicit in the “broad-based growth” discussions that pervaded the 1990 World Development Report (World Bank 1990) and the focus over the past few decades on measuring development progress on the basis of a broader set of indicators related to human development (not just income). Following concerns about the unequal impacts of growth and in the context of explicit global commitments to poverty reduction in the Millennium Development Goals, a new strand of work in the 2000s recatalyzed the discussion on who should benefit from growth. An aca- demic debate on the conceptualization and operationalization of pro-poor growth emerged, with various prevailing views. The “absolute approach,” suggested by Ravallion and Chen (2003) and Kraay (2006), defines any poverty-reducing episode as being pro-poor. The “relative view,” held by Kakwani and Pernia (2000), Son (2004), Klasen (2004, 2008), Essama- Nssah and Lambert (2009), and Negre (2010), requires the poor to benefit disproportionately from growth. In another strand of work, Subramanian (2011) points to the notion of “egalitarian growth,” which requires that at least 40 percent of increases in GDP accrue to the bottom 40 percent. Depending on initial inequality, pro-poor growth could require the growth in incomes of the bottom 40 percent to exceed dramatically the growth in incomes of the overall population. These concepts gave rise to several indicators and measures, with no overall agreement on how to balance the trade-offs in adopting a single measure. Basu (2001, 2006) takes a step further by providing some practical suggestions on how to go about measuring an inclusive growth concept in a systematic way at the global level. He argues that development goals that go beyond income growth to broader objectives—a better quality of 76 UNDERSTANDING SHARED PROSPERITY life, increased education, and a more equitable distribution of goods and services—are indeed desirable. Basu notes that a meaningful summary measure that would capture these multiple objectives is urgently needed. By reasoning that a perfect measure that encompasses all the desired properties does not exist, Basu concludes that a careful balance between conceptual coherence, empirical tractability, and ease of communication needs to be considered. He proposes to focus on the per capita income of the poorest 20 percent of the population and, specifically, the growth rate of the per capita income of the poorest people. Basu further shows how the quintile income has many attractive properties, among them the fact that it correlates more strongly than average per capita income with other (nonmonetary) indica- tors of well-being, such as greater life expectancy and higher literacy. Although this brief description of the evolution of the discussions around the concept of shared prosperity does not aim to be comprehensive, three main messages emerge. First, shared prosperity is a concept that has been discussed, debated, and of concern to the development community for many decades. Second, for any measure of shared prosperity to be of meaningful use, it needs to reconcile multiple challenges, both technical and practical. And finally, the adoption of the goal of achieving shared prosperity inherently implies that ensuring the well-being of the most vulnerable in a society is a key goal of development. Shared prosperity decoded The shared prosperity goal adopted by the World Bank is to boost the per capita income or consumption growth of the poorest 40 percent in a given country.1 In the debates leading to the adoption of this indicator, three aspects influenced the discussion and final choice of the indicator: its simplicity, target population, and theoretical considerations. These are described below. Simplicity An attractive feature of the shared prosperity goal is its conceptual simplic- ity. To track shared prosperity, the only requirement is data on the income (or consumption) growth rate of those in the bottom 40 percent of the population between two periods. The development community has used and will continue to use national GDP growth rates to track the overall 77 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY progress of a given country over time; given the appropriate data, the calculation of the income growth of the bottom part of the distribution is an easy extension. However, there are two important differences. First, income growth in the shared prosperity measurement refers to growth in mean income or consumption of the bottom 40 percent of the population from household surveys.2 As chapter 6 discusses, a key difference between the shared prosperity measure and GDP growth is that living standards measured from surveys and those from national accounts are often strik- ingly different. In this sense, the two growth measures are not comparable. Second, shared prosperity in this framework does not track the growth rate of the same bottom 40 percent of people over time. The shared prosper- ity indicator is an “anonymous” measure, expressed as the growth in mean income of the bottom 40 percent of the income distribution between two periods, irrespective of the individuals belonging to this group at either point in time. It is likely that many of the people in the bottom 40 percent of the income distribution in the first period will not be in the bottom 40 per- cent in the second period. This highlights an important distinction between shared prosperity measured with cross-sectional data at two points in time and the concept of mobility. The latter concept explicitly focuses on track- ing the same individuals (nonanonymous) over time and analyzing their income growth over time (upward or downward), which requires panel data. Unlike the goal to end extreme poverty, the shared prosperity goal is country specific. Therefore, tracking progress only requires monitoring the income growth of the bottom 40 percent over time in a given country. Since it is country specific, there is no explicit target set at the global level. Even at the country level, the shared prosperity goal is more of a moving target than an absolute target. Although the goal of minimizing extreme poverty has a specific target, the goal of boosting shared prosperity does not have a specific numerical target. Shared prosperity is “unbounded” in this way: there is no absolute standard that every country should reach. Even for a given country, the standard is unlikely to remain constant, as societies, along with people’s aspirations, evolve along the development path. Target population The shared prosperity goal’s emphasis on the bottom 40 percent focuses the goal on an explicit objective population. The goal emphasizes that, within a country, different subpopulations will be able to take advantage of economic opportunities in different ways. The focus on growth of the bottom 40 percent ensures that, irrespective of what is happening in the 78 UNDERSTANDING SHARED PROSPERITY country overall, economic growth should also reach the least well-off people in the population. In this way, the focus provides guidance for policy design. Policy design must directly consider the potential impact of policies on the bottom 40 percent and how those at the bottom can benefit the most from the policies. Box 2.2 discusses considerations underlying the choice of 40 percent as the cutoff for defining the shared prosperity goal. Yet, in sharp contrast to Rawls, who emphasizes the poorest person, the shared prosperity indicator is in fact not an egalitarian measure. Although the prosperity indicator does not put any weight on people above the bottom 40 percent, it gives more weight to the richest person within the bottom 40 percent since, by construct, increased shared prosperity can be achieved more quickly by first raising the incomes of those at the top of the bottom 40 percent (chapter 3 discusses this in more detail). Box 2.2 Why 40 percent? The decision to have a shared prosperity goal that population.a The choice of the 40th percentile as focuses on a specific target population required a the cutoff point to define the “least well-off” part of choice for what cutoff to use for the target popu- the population is admittedly a somewhat arbitrary lation. The main arguments given for the use of threshold, although the criticism of arbitrariness 40 percent as the cutoff relate to practical com- would apply to any such threshold. promises regarding trade-offs in the empirical Interestingly, the threshold of 40 percent may implementation of the goal. On the one hand, have some separate empirical relevance. Palma placing the threshold “too high” could result in (2011) explores recent trends in distributional mean per capita (household survey) incomes of the income disparities within countries and finds target population that are very close to national two stylized facts. First, the disparity between the mean per capita incomes (also survey based) and income shares of the top 10 percent of the popula- hence provide little information beyond GDP per tion and the bottom 40 percent has been increas- capita. On the other hand, in many low-income ing over time. At the same time, the income share countries, extreme poverty is concentrated in the of the fifth to ninth deciles has remained roughly bottom quintile, so placing the threshold at 20 per- constant. Palma argues that half the world’s popu- cent would provide little information beyond that lation has acquired strong “property rights” over provided by the extreme poverty goal. Placing the their respective share of national income, while the threshold “too low” could also be problematic if other half of the income is increasingly up for grabs income data on the poorest people, being at the tail between the very rich and the bottom 40 percent. of the distribution, tended to have higher measure- This narrative provides some further rationale for ment errors. The very poor often have no steady focusing specifically on the incomes of the bottom source of income and the income they do have 40 percent. comes from multiple, informal sources that are not a. Note that the frequently asserted claim that measurement always easy to document, which could contribute error is highest for the poorest has not been comprehensively to higher measurement error for this part of the substantiated empirically. 79 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Theoretical considerations Beyond simplicity and focus, it is also desirable for any measure of living standards to satisfy some of the standard axioms of welfare functions. It is worthwhile to mention briefly two of the axioms.3 First, the shared prosperity measure satisfies the criterion of anonymity: in a given society, it should not matter who is at the bottom (or the top) of the income distribu- tion. For the shared prosperity measure, all that matters is the growth rate of the incomes of the bottom 40 percent, irrespective of who they may be. The shared prosperity indicator also satisfies the weak Pareto principle, which states that if the income of every individual in a group rises, the group is considered better off. Since the aim of the shared prosperity goal is to track the income growth of the bottom 40 percent of the population, while the income of those outside the range is not relevant for the indica- tor, an increase in the incomes of the bottom 40 percent will by design improve shared prosperity and hence be interpreted as an improvement in the well-being of the group. No welfare index is perfect, and the shared prosperity indicator is not an exception. For example, the shared prosperity indicator does not satisfy the weak transfer axiom, which requires lump-sum transfers from a richer person to a poorer one within a group to lead to improvement in the value of the indicator. Since the shared prosperity indicator focuses on the bot- tom 40 percent and does not differentiate within this group, such a transfer would not affect shared prosperity. The weak transfer axiom would likely be less of an issue for a cutoff point of, say, 20 percent, highlighting one of the trade-offs in the choice of the threshold, as discussed in box 2.2. An interesting implication of the shared prosperity indicator is that any change in the distribution of income within the bottom 40 percent that maintains the same total share of income accruing to the bottom 40 percent would not affect the value of the measure. This aspect of the indicator is discussed further below. Tracking shared prosperity in practice Who are in the bottom 40 percent? As is the case with the measurement of poverty, an important entry point for reflecting on government action and policy on shared prosperity relates to the profile of the bottom 40 percent. In what way do the characteristics 80 UNDERSTANDING SHARED PROSPERITY of the population in the bottom 40 percent of a given country differ from those of the population as a whole (or the top 60 percent or the poor)? Since shared prosperity is a relative concept, it corresponds to different income groups across countries. In other words, the bottom 40 percent differs across countries. For example, the average household in the bot- tom 40 percent of the income distribution in the United States would be among the richest 10 percent in Brazil (figure 2.1). Similarly, the average household in the bottom 40 percent of Brazil’s income distribution would be at approximately the 90th percentile of the income distribution in India. Not only does the average income of the bottom 40 percent differ across countries, but the composition of incomes among the bottom 40 percent is also likely to vary substantially. Figure 2.2 plots the sizes of various income-based groups across a selection of countries. The groups are the Figure 2.1 The bottom 40 percent in the United States, Brazil, and India, 2008 64,000 32,000 Income or consumption, per capita 16,000 (2005 PPP, log-scale) 8,000 4,000 2,000 1,000 500 250 1 2 3 4 5 6 7 8 9 10 Decile of national income or consumption distribution United States Bottom 40% mean Brazil Bottom 40% mean India Bottom 40% mean Source: Based on Lakner and Milanovic (2013). Note: The lines connect decile means for each country, where the thicker portion of each line connects the decile means of the bottom 40 percent. PPP = purchasing power parity. 81 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 2.2 The bottom 40 percent can encompass various income groups across countries, circa 2009 Russian Federation Thailand Tunisia Turkey Chile Armenia Costa Rica Peru Kyrgyz Republic Brazil China South Africa Honduras Ethiopia India Lao PDR Bangladesh Angola Mali 0 20 40 60 80 100 Population share (percent) Extreme poor (less than $1.25 a day) Moderate poor ($1.25 to $4 a day) Vulnerable ($4 to $10 a day) Middle class and rich (more than $10 a day) Source: Based on data for latest year available from the World Bank PovcalNet database (accessed August 2014). Note: The groups in the figure are the extreme poor, as defined by the World Bank’s international poverty line; the moderate poor, who live on between $1.25 and $4 a day; the vulnerable who live on between $4 and $10 a day; and the middle class and rich, who live on more than $10 a day—all measured at 2005 constant purchasing power parity (PPP). The concept of people living on between $4 and $10 a day being considered vulnerable is based on evidence that a considerable share of households above a given poverty line is usually vulnerable to falling below that line over time. See Ferreira and others (2012) and Birdsall, Lustig, and Meyer (2014). The vertical line is drawn to show the proportion of the population living in the bottom 40 percent. extreme poor, as defined by the World Bank’s international poverty line (less than $1.25 a day); the moderate poor, who live on between $1.25 and $4 a day; the vulnerable, who live on between $4 and $10 a day; and the middle class and rich, who live on more than $10 a day, all measured at 2005 constant purchasing power parity (PPP).4 In countries such as Angola, Bangladesh, and Mali, the bottom 40 percent essentially captures 82 UNDERSTANDING SHARED PROSPERITY the extreme poor (using the international poverty line). In Ethiopia and India, 80 percent of those in the bottom 40 percent are extremely poor and the rest are moderately poor. In other countries, such as China, the bottom 40 percent predominantly comes from the moderately poor (with the rest extremely poor). A different picture emerges in countries in Latin America and the Caribbean and in Europe and Central Asia. In some of the richer, upper- middle-income countries in these regions, such as Chile and the Russian Federation, the large majority of individuals in the bottom 40 percent are in the vulnerable group. The vulnerable are nonpoor individuals with a high risk of falling back into poverty. In all, these trends highlight the great range of people that constitute the bottom 40 percent across the world. Consequently, the concept of shared prosperity and its associated policy discussion will have different meanings in each country.5 Since the concept of shared prosperity focuses attention on the poorest members of society within a country, it is useful to compare the bottom 40 percent with the poor as defined by each country’s respective national poverty line. In figure 2.3, data on national poverty lines from the data set compiled by Ravallion, Chen, and Sangraula (2009) are matched with corresponding surveys from the PovcalNet database. In some cases, focus- ing on the bottom 40 percent captures a narrower group than those living below national poverty lines. For example, in Colombia, Georgia, the Republic of Congo, República Bolivariana de Venezuela, Tajikistan, and The Gambia, the bottom 40 percent implies a much narrower focus on the poor than the national poverty line. By contrast, in China, India, and Tunisia, the concept of shared prosperity encompasses a much larger part of the population than the poor as defined by the national poverty line. Overall, the comparison of who constitutes the bottom 40 percent across countries illustrates the benefit of a “moving target” that can be interpreted differently in different countries, rather than a common goal for all countries. In some cases, especially in low- and lower-middle-income countries, the shared prosperity indicator can reinforce national poverty lines and strengthen the focus on the poor. In other cases, especially in upper-middle-income countries, the shared prosperity indicator may help broaden the policy agenda for nonpoor parts of the population that might otherwise face the risk of being left behind or falling into poverty. In the latter case, policy agendas for the bottom 40 percent may introduce trade- offs or complementarities with those for the poor. Although a discussion of this type of trade-off is beyond the scope of this report, it is worth noting 83 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 2.3 The bottom 40 percent compared to the poor as defined by national poverty lines Tunisia 1995 China 2002 Estonia 1995 India 2004 Jordan 2002 Bulgaria 2001 Morocco 1998 Chile 2000 Bosnia and Herzegovina 2001 Sri Lanka 2002 Latvia 1995 Albania 2002 Thailand 1992 Russian Federation 2002 Vietnam 2002 Peru 2000 Burkina Faso 2003 Indonesia 1999 Turkey 2002 Senegal 1991 Côte d’Ivoire 1998 Cameroon 2001 Yemen, Rep. 1998 Kazakhstan 1996 Philippines 1988 Bangladesh 2000 Cambodia 2004 Guinea-Bissau 1991 Azerbaijan 2001 Mauritania 2000 Pakistan 1998 Moldova 2001 Belarus 2002 Mexico 2002 Kenya 1997 Armenia 1998 Ethiopia 1999 Mozambique 2002 Bolivia 2001 Georgia 1997 Djibouti 2002 Paraguay 2002 Venezuela, RB 1989 Gambia, The 1998 Colombia 1999 Congo, Rep. 2005 Tajikistan 1999 0 20 40 60 80 Population share (percent) Sources: Based on data from the World Bank PovcalNet database (accessed August 2014) and Ravallion, Chen, and Sangraula (2009). Note: The blue bars show the percentage of the population below the national poverty line. The vertical line is drawn to show the proportion of those in the bottom 40 percent who are also living below the national poverty line. 84 UNDERSTANDING SHARED PROSPERITY that such trade-offs are likely to be a country-specific issue and a function of the connections between growth and poverty reduction in specific con- texts, which are discussed further below and in chapter 3.6 Measurement of shared prosperity at the country level: What constitutes success? Calculating shared prosperity requires measuring income or consumption in a given country and year for those in the bottom 40 percent. Then, by replicating this calculation in a comparable manner in another year, shared prosperity in a particular country can be measured as the annu- alized growth rate between the two periods (box 2.3 discusses the data requirements). What, then, constitutes success? A positive number would suggest that the bottom 40 percent of the population saw increases in their incomes, but, given the fact that the goal is unbounded, the number would not provide meaningful information about how good the growth rate was. This section discusses some ways to think about assessing the performance of the shared prosperity goal. A first insight about a country’s performance can be inferred by having multiple periods of data, so that the evolution of shared prosperity can be tracked over time. Figure 2.4 shows examples from four countries where multiple surveys exist from the 1980s onward. In Sri Lanka, shared pros- perity increased slowly in the late 1980s and early 1990s before increasing more sharply in the 2000s. Similarly, in Brazil, shared prosperity fluctu- ated in the 1990s before beginning a strongly increasing trend from the early 2000s. In this sense, performance has been better in both countries in more recent years. This is also the case in South Africa and Uganda. Another possibility is to compare the performance of the bottom 40 per- cent with that of other parts of the income distribution (for example, the top 60 percent of the population) or overall national average performance. Alongside trends in the average income of the bottom 40 percent, figure 2.4 also shows annualized growth rates for the population as a whole. In addition to providing a means to compare the performance of shared pros- perity across countries, this comparison also allows an assessment of the evolution of income inequality (this point is discussed further below). For example, the bottom 40 percent in South Africa did better than average during the mid-1990s (suggesting not only that incomes of the bottom 40 percent grew, but also that there was some catching up). By contrast, by the 2000s, income growth for the bottom 40 percent increased compared with 85 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Box 2.3 Measuring and tracking shared prosperity at the country level Measurement of the shared prosperity indicator, As discussed in chapter 1, the use of income like any measure, implies information requirements or consumption to measure poverty has a long and potentially some underlying assumptions. tradition, although consumption is usually the Chapter 1 discusses some of the issues related to preferred indicator, particularly in developing the measurement of poverty, this chapter discusses countries. As with the measurement of poverty, shared prosperity, and chapter 5 discusses addi- data on consumption, if available, are preferred tional data challenges and promising areas that can over data on income to construct the shared pros- improve measurement for both goals. perity indicator. Conceptually, consumption is usually less susceptible to measurement errors and What data to use? temporary fluctuations and thus it is often seen as The shared prosperity measure refers to growth a better measure of current living standards.b Even of the per capita real household income or con- poor households can usually rely on some form sumption of the bottom 40 percent of a population. of saving and dissaving mechanisms to smooth Although aggregate growth statistics like GDP income shocks. Consumption is usually also seen as more indicative of long-term living standards.c growth are typically derived from national income Empirically, the difference between a shared pros- accounts, these aggregate data cannot provide dis- perity measure based on consumption and one aggregated information on the growth of income based on income is not trivial (as figure 2.5 shows). fractiles, such as the bottom 40 percent. Instead, shared prosperity can be calculated from nationally Which time interval to consider? representative surveys that provide income or con- In many countries, data availability will dictate sumption data at the household level. These surveys the interval in which shared prosperity can be directly allow the identification of the bottom 40 estimated. But in cases where data from multiple percent of the population and their incomes.a years are available, how should the shared prosper- ity measure be calculated? When there are more Which measure of well-being? data, the best approach is to take advantage of all To construct the shared prosperity indicator with the information available (see fi gure 2.6 for an at least two data points for a given country over illustration). time, strict comparability is required between the a. Note that while these surveys are well designed to cap- respective surveys used. Constructing a growth rate ture population averages, it is still the case that they may not with consumption levels in one year and income adequately capture information from specific subpopulations in the other year is a meaningless option. Even where extreme poverty is concentrated. b. There is a large literature on the use of income or con- when the same indicator is used, special attention sumption in the measurement of poverty. Ravallion (1994), is needed to ensure that the methodology used to Deaton (1997), and Deaton and Zaidi (2002) provide a construct the measure of well-being is consistent in comprehensive overview of the conceptual background and the two years. Many countries periodically update empirical issues. For a broader review of the literature on living standards measurement, see Slesnick (1998). their survey methodologies to account for changes c. In economic theory, the permanent income hypothesis in population structures, spatial prices, caloric essentially states that individuals base their consumption requirements, and imputation techniques—adding decisions on their anticipated long-term income rather than new sources of consumption or income, treatment shorter-term fluctuations. However, Deaton concludes that “the standard argument—that by the permanent income hypothesis, of taxes, and food consumed away from home, to consumption is a better measure of lifetime living standards name a few. Growth rates can be sensitive to such than is current income—is much weaker than the arguments changes and thus comparability is essential. based on practicality and data” (Deaton 1997, 148). 86 UNDERSTANDING SHARED PROSPERITY Figure 2.4 Evolution of mean income or consumption of the bottom 40 percent and the overall population, 1980–2010 a. Brazil b. South Africa Cumulative growth (baseline 1993) (percent) Cumulative growth (baseline 1992) (percent) 120 120 100 100 80 80 60 60 40 40 20 20 0 0 1980 1990 2000 2010 1980 1990 2000 2010 Year Year Cumulative growth (baseline 1985) (percent) Cumulative growth (baseline 1989) (percent) c. Sri Lanka d. Uganda 120 120 100 100 80 80 60 60 40 40 20 20 0 0 1980 1990 2000 2010 1980 1990 2000 2010 Year Year Mean income or consumption of the total population Mean income or consumption of the bottom 40 percent Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: Cumulative growth of household consumption expenditure or income per capita in constant 2005 purchasing power parity (PPP). the mid-1990s but was significantly slower than average income growth, implying increased inequality. Although shared prosperity was boosted over this period in South Africa (average incomes of the bottom 40 percent increased), the bottom 40 percent underperformed relative to the rest of the population. The same is true in Sri Lanka in the 1990s and 2000s. In Uganda, the trends suggest not only that shared prosperity has been 87 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY increasing over time, but the bottom 40 percent also did at least as well as the rest of the population (the growth rate was the same as or exceeded the overall average). The same is true for the bottom 40 percent in Brazil in the second half of the 2000s. The aggregate shared prosperity indicator provides no information about the performance within the bottom 40 percent. As is the case for the measurement of poverty, or indeed any other indicator of develop- ment outcomes, subnational divergence may confound the interpretation of shared prosperity. The perceived performance of shared prosperity can vary depending on whether the income or consumption growth of the bottom 40 percent is calculated at the level of the country, state, or even district—particularly in large and heterogeneous countries. This problem can arise through a scale effect, where the level of aggregation can deter- mine the findings for a specific area (for example, state averages can smooth out district-level heterogeneity), and a zoning effect, where the choice of groupings of households at the same scale can similarly influence results.7 The challenge of interpreting shared prosperity at the subnational level is illustrated by the case of rural India, where regional disparities in poverty and other development indicators have traditionally been large, between and within states. In figure 2.5, shared prosperity using growth in household monthly per capita expenditure for rural India is calculated for the bottom 40 percent at the national level, the state level, and the district level. At the national level, the mean household per capita expen- diture of the rural bottom 40 percent fell by about 4 percent (red dashed line in both panels). Despite the weak performance during this period, these results hide the wide heterogeneous performance at the subnational level. For example, the mean household per capita expenditure of the rural bottom 40 percent in Rajasthan grew by 4.4 percent, while it fell by 6 percent in neighboring Gujarat (figure 2.5, panel a). Similar trends can be seen within states: within rural Gujarat, the state’s poor performance masks the fact that the mean of the bottom 40 percent grew by 2 percent in the state’s most populous district of Ahmedabad, while it fell in most other districts (panel b). Thus, the overall decrease in shared prosperity at the national level masks mixed performances within states and districts. The source of data or time interval chosen can also affect the perfor- mance and interpretation of the shared prosperity indicator. Since Peru collects annual data on consumption and income, it is a natural candidate to examine sensitivity to the source of data and time period used. Figure 2.6 shows the growth rates for the bottom 40 percent measured with 88 UNDERSTANDING SHARED PROSPERITY Figure 2.5 Shared prosperity in rural India at various levels of disaggregation, 2007/08–2009/10 a. By state b. By district Lakshadweep Nagaland Dadra and Nagar Haveli Mizoram Meghalaya Bihar Madhya Pradesh Assam Karnataka Jharkhand Manipur Chhattisgarh Gujarat Tamilnadu Kerala Uttar Pradesh Orissa Himachal Pradesh West Bengal Tripura Delhi Sikkim Daman and Diu Maharashtra Andhra Pradesh Punjab Uttaranchal Haryana Rajasthan Andaman and Nicobar Islands Pondicherry Goa Chandigarh –0.2 0 0.2 0.4 0.6 0.8 –0.5 0 0.5 1.0 1.5 Growth of the bottom 40 percent Growth of the bottom 40 percent Source: Based on data from India’s NSSO National Sample Survey rounds 64 (2007/08) and 66 (2009/10). Note: Orange dashed line marks the annualized growth in consumption expenditure of the rural bottom 40 percent for the nation. Growth rates calculated using real household monthly per capita expenditure (mixed reference period), deflated using the consumer price index (CPI) for agricultural laborers with base 1986/87 = 100. income and consumption, using different time intervals. The end year is fixed at 2012, but the initial year from which the annualized growth rate is estimated ranges from 2011 to 2004. A few results stand out. First, the shared prosperity indicator for a given interval with consumption data is different (in most cases lower) than the one with income data. The growth rates can differ by as much as 2 percentage points (about 30 percent), thus providing substantially different interpretations about the performance of Peru. Even more striking, the choice of time interval used to calculate shared prosperity makes a significant difference to Peru’s performance. At the extreme, shared prosperity (growth in mean consumption of the 89 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 2.6 Illustration of how the choice of data and time interval influence shared prosperity estimates 8 7 Average annual growth (percent) 6 5 4 3 2 1 0 2 2 2 2 2 2 2 2 –1 –1 –1 –1 –1 –1 –1 –1 04 05 06 07 08 09 10 11 20 20 20 20 20 20 20 20 Consumption expenditure Income Source: Based on data from Peru’s Encuesta Nacional de Hogares (ENAHO) household survey. bottom 40 percent) between 2011 and 2012 is 6 percent, while between 2010 and 2012 (just adding one more year), it is around 1 percent. Similar points can be made when deciding which start and end dates to use. Instead of holding the end year constant and varying the time inter- val, as in figure 2.6, the end years in figure 2.7 vary to compose estimates of equal duration over a different range of years. Once again, the results vary when consumption or income data are used to estimate the shared prosperity indicator. In addition, the choice of which five-year interval one uses affects the performance of shared prosperity. For the five-year span between 2004 and 2008, income growth of the bottom 40 percent in Peru is estimated at 6 percent (annualized) compared with 10 percent if the period used shifts by one year (2005 to 2009). Given the sensitivity of estimates to the source of data and time intervals used, a last exercise is to consider how the performance of shared prosperity fares by estimating moving averages of income and consumption growth (as depicted in figure 2.7). The results strongly show how both two- and three-period moving averages smooth considerably the estimated shared prosperity. 90 UNDERSTANDING SHARED PROSPERITY Figure 2.7 Moving averages provide more stable shared prosperity estimates 12 Average annual growth (percent) 10 Three-period moving 8 average of consumption expenditure 6 Two-period moving average of consumption expenditure 4 2 2004–08 2005–09 2006–10 2007–11 2008–12 Consumption expenditure Income Source: Based on data from Peru’s Encuesta Nacional de Hogares (ENAHO) household survey. A take-away message from this section is that in the presence of more rounds of comparable data, they should be used to improve the interpreta- tion and better assess the performance of a country in boosting shared pros- perity. Moving averages provide a simple and effective way to smooth some of the inherent variations observed in the underlying data and thereby pro- duce more stable results. In most countries, annual consumption or income data are not consistently available. The case of Peru, used to illustrate the scenarios above, is an outlier in terms of good data availability. Although the presence of good data in Peru is closer to the situation of some more devel- oped countries, it is not the norm in the majority of developing countries (see chapter 5 for a more detailed discussion of data availability across countries). Still, Peru showcases a second take-away and an aspirational scenario that other countries will need to move to, so that frequent data can better accom- modate measuring and tracking development goals like shared prosperity. Monitoring performance in a global context Shared prosperity is a country-specific goal that can help a given country better understand whether growth is benefiting those at the bottom 40 percent of the population. Still, monitoring of shared prosperity could also involve country-by-country comparisons, which may reveal some 91 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 2.8 Shared prosperity, by country Bolivia Tanzania Russian Federation Cambodia Belarus Colombia Slovak Republic Uruguay Peru Paraguay Congo, Rep. Nepal China Argentina Bhutan Vietnam Brazil Kyrgyz Republic Moldova Costa Rica Panama Turkey Botswana Ukraine Nicaragua Rwanda Ecuador Thailand South Africa Honduras Estonia Chile Mozambique Uganda Tunisia Namibia India Poland Pakistan Sri Lanka Jordan Montenegro Dominican Republic Mali West Bank and Gaza Czech Republic Bangladesh Slovenia Lao PDR Bulgaria Philippines El Salvador Lithuania Georgia Armenia Mexico Latvia Iraq Mauritius Senegal Nigeria Ethiopia Hungary Albania Malawi Serbia Guatemala Togo Madagascar −10 −5 0 5 10 15 Annualized income or consumption growth rate of the bottom 40 percent (percent) Middle East & North Africa Latin America & Caribbean South Asia East Asia & Pacific Europe & Central Asia Sub-Saharan Africa Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: Growth rates in shared prosperity are calculated as annualized growth rates in per capita income or consumption expenditure over the period 2006–11. Given the data available from PovcalNet for each country, surveys that use the same welfare indicator are matched as closely as possible to this period. 92 UNDERSTANDING SHARED PROSPERITY interesting regional or global trends in how countries are performing with respect to growth of the income of the bottom 40 percent of their popu- lations. Figure 2.8 presents a summary of comparable shared prosperity performances for 69 countries for the period 2006 to 2011.8 Overall, the bottom 40 percent has performed well. With the exception of 11 of the countries of this sample, the bottom 40 percent in all countries experi- enced increases in their incomes during this period. However, there is considerable variation across countries. Countries such as Belarus, Bolivia, Colombia, Russia, Tanzania, and Uruguay have experienced annual growth rates in income or consumption of the bottom 40 percent of more than 8 percent, while countries such as Albania, Guatemala, Madagascar, Malawi, Serbia, and Togo have experienced negative annual growth. In addition, a more nuanced understanding of the comparison may result from considering the drivers of shared prosperity growth. The cases of Honduras and Vietnam help illustrate this point. Over approximately the same period, the growth rate of the bottom 40 percent in Honduras was 4 percent (per year), while in Vietnam it was 6 percent. Does that mean that Vietnam did better at boosting shared prosperity during this period? A strict interpretation of the shared prosperity measure would imply that this was the case. Still, overall growth in Honduras for the whole popula- tion during this period was 2 percent, while in Vietnam it was 8 percent. Therefore, in Honduras, income growth of the bottom 40 percent was almost double the national average. In Vietnam, despite the gains for the bottom 40 percent and a high overall growth rate, growth of the bottom 40 percent has lagged behind the national average. Given the discussion on how shared prosperity is a moving target, as well as its inherent connection with inequality, this kind of nuance is worth keeping in mind when com- paring trends between countries. Box 2.4 offers some additional discussion and suggestions on cross-country comparisons of shared prosperity. The “sharing” in shared prosperity To what extent does “sharing” prosperity imply the bottom 40 percent of the population should have a larger slice of the pie, beyond simply sharing in the proceeds of growth? The shared prosperity indicator on its own is not an inequality measure; it is a simple growth measure that tracks how the bottom 40 percent of a given country are doing, without any need to compare this progress with other parts of the population. Still, comparing the growth of the bottom 40 percent to the average for the total population 93 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Box 2.4 The challenges of measuring and tracking shared prosperity at the global level Cross-country comparisons and tracking of shared countries of household survey data. Furthermore, prosperity pose several additional challenges to although poverty historically has been tracked only monitoring shared prosperity at the country level, for developing countries, shared prosperity aims to since to create a global cross-country data set, provide measures in all countries across the world comparability across various dimensions needs to (data permitting), adding to the harmonizing com- be met. Many of these issues are similar to those plexity in the scope as well as the technical differ- when comparing poverty numbers across countries, ences involved. The choice of the time period over although the dynamic aspect of the shared prosper- which to assess growth can also matter significantly ity indicator brings some additional technical obsta- at the global level: electoral cycles, country-specific cles. The main message below is that, despite the economic upswings or downturns, and natural disas- best efforts and intentions, comparisons of shared ters can all affect individual country performance prosperity across countries should be interpreted and thereby render comparisons difficult to make. with considerable caution. Data. The primary data source for cross-country Which measure of well-being? As with the country comparisons of shared prosperity is the World level, strict comparability of the well-being indicator Bank’s PovcalNet database. It comprises income being used is desirable to make comparisons across or consumption aggregates from household sur- countries. In practice, consumption data are not veys conducted by statistical offices of individual always available, and in recent years many countries countries and harmonized to achieve some degree (especially middle-income countries) have tended of comparability across countries and years. Since to use income to measure well-being. Following PovcalNet is used for estimating poverty rates at strict requirements of comparability can seriously international poverty lines, using it to estimate restrict the scope of any exercise to compare shared shared prosperity can also have the advantage of prosperity across countries, but even when the same ensuring the consistency of shared prosperity num- measure of well-being is used, there can still be bers with country-level poverty trends. One limita- important differences in the quality of data and tion is that PovcalNet does not include information their comparability across countries and over time. from advanced countries, implying that compari- Most of these issues are a direct or indirect result sons that aim to present estimates for all countries of variability in how data on different categories of around the world will require a blend of data sets, expenditures or income are collected and treated in rendering comparability issues more difficult. household surveys across countries and over time. Imputing or using survey data? One way to compare Timing—which period and how often to update? shared prosperity across countries with PovcalNet Cross-country comparisons will involve fi xing a data is to base the comparison on a common base specific common time interval in all countries over year —an extension of the approach used to measure which to assess shared prosperity. Since household extreme poverty across countries discussed in chap- surveys are infrequent in most countries and mis- ter 1. This requires that country estimates are “lined aligned in terms of their timing, perfect compara- up” first to a common reference year, interpolating bility is impossible without drastically reducing the for countries in which survey data are not avail- number of countries for which the indicator can be able in the reference year but are available either reported. Countries do not generally coordinate the before or after, or both. The more survey data are fielding of household surveys, and, as a result, at any available for different years, the more accurate the point in time there will only be partial coverage across (continued) 94 UNDERSTANDING SHARED PROSPERITY Box 2.4 Continued interpolation. This approach has the advantage of not the analyst wants to be in terms of the surveys aligning shared prosperity estimates to those for being close to the interval of interest. These are again poverty, so that monitoring of both goals can pro- not trivial decisions, as has been shown through- vide contemporaneous information and insights. out this chapter, since all three choices will affect By lining up the year of comparison, this approach inferences about performance for each country in would also potentially make it possible to construct specific (and different) ways, making comparability regional shared prosperity aggregates, which could complicated. As an example, in a recent application shed light on more general aggregate patterns of of this approach, Narayan, Yoshida, and Mistiaen growth for the bottom 40 percent. (2013) chose the time interval to be the five-year Unfortunately, to line up countries in a common period between 2005 and 2010. For the choice of year, the process of interpolation requires adjusting living standards measure, the authors use annualized the mean income or consumption observed in the average growth rates in per capita real income and survey year by a growth factor to infer the unob- consumption, which has the advantage of increas- served level in the reference year. In practice, since ing the number of countries in the pooled sample. survey data in most countries are not available on Finally, growth rates for the bottom 40 percent are an annual basis, the change in private consumption calculated only for those countries meeting the fol- per capita as measured from the national accounts is lowing survey criteria: (a) the latest household survey used to calculate this growth factor. There can be no for a given country is no older than 2008; (b) the guarantee that the survey-based measures of income survey year for the initial period is close to (t 1 − 5) or consumption change at exactly the same rate within a bandwidth of ± 2 years; and (c) living stan- as private consumption in the national accounts, dards aggregates (consumption or income) for both even if this appears to be the best currently avail- years within a country are comparable. Based on able option for global poverty estimates. Chapter 6 these rules, the final set of countries only covers 79 of discusses this further and provides examples of this a potential 150 countries in PovcalNet, comprised to problem in practice. Perhaps more troublesome, this a large extent of countries in Latin America and the calculation rests on the assumption of distribution- Caribbean and Eastern Europe and Central Asia. ally neutral growth: income or expenditure levels What is the take-away message of all this? First, are adjusted for growth between periods assuming use of actual surveys is more sensible for estimating that the underlying relative distribution of income shared prosperity. Second, whatever assumptions or expenditure observed in survey years remains can be made from the use of existing surveys, mak- unchanged. This assumption is problematic since ing cross-country comparisons on the performance changes in the underlying relative distribution of of shared prosperity should be done with consider- income or expenditure are exactly what the shared able caution, since the results for a given perfor- prosperity indicator seeks to capture. mance can differ depending on which rules are An alternative approach is to use actual surveys used. In addition, although the notion of an aggre- around a fixed interval . The large benefit of this gate global or regional measure of shared prosperity approach is that it uses actual surveys as opposed to is indeed appealing, given the data limitations, its projections for those countries and years where data interpretation would be misleading. Whatever deci- do not exist for the reference year. This approach sions are made, shared prosperity should remain requires a decision on the interval of interest; a country-specific goal, and each country should whether consumption, income, or both can be used decide on the specific goals and metrics by which (for different countries); and how conservative or to monitor its performance over time. 95 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY can provide a gauge as to whether people at the bottom are doing better or worse than the average (or, in practice, the top 60 percent of the popula- tion). Moreover, changes in shared prosperity can be driven as much by changes in inequality as by changes in national growth. A nice way to illustrate this is to decompose the change in shared prosperity between two years into two parts: the change in the average income of the bottom 40 percent is a combination of change in the share of the total income accruing to the poorest 40 percent and change in the average income for the population as a whole—in other words, differences in how much of the total pie the bottom 40 percent has managed to accrue and how quickly the overall pie has grown.9 Figure 2.9 shows an approximation of this decomposition of the shared prosperity indicator across 69 countries from around 2006 to 2011. The graph plots the annual growth in household survey mean income or consumption expenditure against the annual growth in the share of total income or consumption accruing to the bottom 40 percent in each Figure 2.9 Growth and changing shares of income 15 Annualized growth in the income or consumption share of the bottom 40 percent (percent) 10 Bolivia Cambodia 5 Mali Ecuador Colombia Dominican Republic South Africa Tanzania 0 Serbia Madagascar India China Nigeria −5 −10 −10 −5 0 5 10 15 Annualized income or consumption growth rate of the total population (percent) Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: Growth rates in shared prosperity are calculated as annualized growth rates in per capita income or consumption expenditure over the period of circa 2006–11. See note to figure 2.8 for further explanations. 96 UNDERSTANDING SHARED PROSPERITY country. The decomposition can illustrate the range of country experiences in terms of the source of the changes in shared prosperity. For example, in Colombia, South Africa, and Tanzania, growth in mean incomes con- tributed a larger part to the overall increase in shared prosperity than the increase in the share of total income accrued by the bottom 40 percent. By contrast, in countries such as Bolivia, Cambodia, and Ecuador, increases in the share of income accruing to the bottom 40 percent was the factor that contributed more to shared prosperity as opposed to the increase in overall mean growth. In countries such as China or India, boosting shared prosperity was a result of high overall growth in mean income, compensat- ing small declines in the share accrued to the bottom 40 percent. Finally, in Madagascar and Serbia, overall shared prosperity decreased, driven by a reduction in both overall mean incomes as well as the share component. In general, these insights showcase how cross-country comparisons should be treated with some caution, as equivalent performances could be the result of different economic processes at play. More conceptually, although policies that promote overall growth can be beneficial for the bottom 40 percent, increased participation of the bottom 40 percent is an alternative way to boost shared prosperity. The shared prosperity goal itself is agnostic on whether changes in shared prosperity should come from changes in growth or changes in inequality. However, this review of past experience shows that different combinations have played a role in boosting shared prosperity in different countries. This is an important aspect of tracking shared prosperity, as it can inform how changes in this mix influence living standards for the bottom 40 percent. Can boosting shared prosperity help end global poverty? As chapter 1 shows, solely accelerating GDP per capita growth rates will not suffice to reduce global poverty to 3 percent by 2030. Is there a link between poverty reduction, shared prosperity, and overall growth? Dollar, Kleineberg, and Kraay (2013) argue that overall income growth accounts for most of the variation in income growth of the bottom 40 percent, thereby suggesting that overall prosperity and shared prosperity are closely related. Skoufias, Tiwari, and Shidiq (2014) also find a strong positive cor- relation between the growth rate of overall consumption and the growth rate of the consumption of the bottom 40 percent across provinces in 97 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Box 2.5 Does inequality affect income growth “equally”? In their paper, van der Weide and Milanovic (2014) associations suggest large income growth effects explore the relationship between initial inequal- among the poor: a one standard deviation reduc- ity and subsequent growth at various parts of the tion of the overall Gini (0.04 points) would imply income distribution. As they note, the fact that a doubling of the growth of the bottom 10 percent many studies do not find a systematic relationship of the population (from an annualized 0.8 percent between inequality and growth could be driven by to 1.7 percent), while the effect of the inequality the simple insight that most studies explore how reduction for the richest 10 percent is muted, at inequality is associated with growth of the aver- only 0.3 percent (from 2.0 percent to 1.7 percent). age income as opposed to growth at different parts Finally, the authors find similar trends when they of the distribution. This is what they set out to separately test whether initial inequality in the explore with data from the United States covering bottom or top 40 percentiles affects subsequent five decades between 1960 and 2010. Specifically, income growth: both top and bottom inequalities they assess the impact of overall inequality, as well are negatively associated with real income growth as inequality among the poor and among the rich, for the poor, while bottom inequality is positively on the growth rates along various percentiles of the associated with the income growth for the rich (no income distribution. association is found between top inequality and Three sets of results stand out. First, among the income for the rich). poor, overall initial inequality is negatively corre- lated with subsequent growth. By contrast, the cor- relation is positive among the rich. Second, these Source: Based on van der Weide and Milanovic (2014). Thailand. They find a significant negative correlation between changes in overall inequality at the province level and the growth rate of consumption of the bottom 40 percent, suggesting that reductions in overall inequal- ity may have an impact on the consumption growth of the poor. Similar results are found in developed settings, as in the case of the United States, in van der Weide and Milanovic (2014; box 2.5). As figure 2.9 indicates, strong overall growth performance is indeed good for the bottom 40 percent. This is confirmed directly in figure 2.10: the same data from around 2006 to 2011 show a positive association between the income growth of the bottom 40 percent and growth in aver- age income. A correlation coefficient of 0.82 suggests that the relationship is strong. However, this correlation also implies that overall growth does not fully explain changes in shared prosperity, which is consistent with the discussion of figure 2.9. In fact, the mean growth rate of mean income of the bottom 40 percent across this sample of 69 countries was 3.6 per- cent, higher than the 2.6 percent per capita income growth of the overall 98 UNDERSTANDING SHARED PROSPERITY Figure 2.10 Shared prosperity and average income growth 15 Annualized income or consumption growth rate Bolivia of the bottom 40 percent (percent) 10 Cambodia Tanzania Colombia China 5 South Africa Ecuador Mali India Dominican Republic 0 Nigeria Serbia Madagascar –5 –10 –10 –5 0 5 10 15 Annualized income or consumption growth rate of the total population (percent) Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: Growth rates in shared prosperity are calculated as annualized growth rates in per capita income or consumption expenditure over the period of circa 2006–11 (all survey based). See note to figure 2.8 for further explanations. population (in line with preliminary estimates done by Narayan, Saavedra- Chanduvi, and Tiwari [2013]). During this period, the bottom 40 percent experienced higher growth rates compared with the overall population in 48 of the 69 countries in the sample. Interestingly, these trends contrast earlier results (for example see Dollar and others 2014) that show a more distribution-neutral nature of growth for similar spells (in length) dur- ing the last few decades. It is not clear to what extent these recent trends represent a departure from the previous patterns and if they will continue. But, the likelihood of achieving the poverty goal is certainly improved if this trend continues, as chapter 1 illustrates that growth alone is unlikely to be sufficient to reach the 3-percent goal. Given the relationship between shared prosperity and growth, how does shared prosperity relate to poverty reduction? A high correlation between shared prosperity and poverty reduction is expected, given that the bot- tom 40 percent of many countries overlaps strongly with those below the poverty line (as shown in figure 2.3). In addition, as chapter 1 and the 99 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY discussion above suggest, both growth and inequality reduction—which are interconnected with shared prosperity—are closely interlinked with poverty reduction. Figure 2.11 shows the annual change in the poverty headcount rate using the $1.25 international poverty line against the annual growth rate of household income or consumption for the total population (panel a) and the bottom 40 percent (panel b) for the same set of 69 countries and time interval (2006 to 2011). The figure differentiates the countries by poverty rate in the initial period, with a larger circle denoting a higher initial pov- erty rate. The correlation between overall income or consumption growth and poverty reduction as well as the correlation between shared prosperity and poverty reduction are strong (and negative) but imperfect. A comparison of the panels can also help explore whether shared prosperity is particularly relevant for poverty reduction. In order to evalu- ate this, two regressions are done: regressing the change in the poverty Figure 2.11 The association of poverty reduction with overall income growth and shared prosperity a. Overall income growth b. Shared prosperity Annual change in poverty headcount at $1.25 a day Madagascar Annual change in poverty headcount at $1.25 a day 2 2 Madagascar Nigeria Nigeria 0 0 Malawi Malawi (percentage points) (percentage points) China −2 Uganda −2 Uganda China India India −4 −4 Cambodia −6 Cambodia Tanzania −6 Tanzania −8 −8 −10 −5 0 5 10 15 −10 −5 0 5 10 15 Annualized income or consumption Annualized income or consumption growth rate of the total population growth rate of the bottom 40 percent (percent) (percent) Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: The size of the circles in the figure denotes poverty headcount rate in the initial period. Growth rates in shared prosperity are calculated as annualized growth rates in per capita income or consumption expenditure over the period of circa 2006–11. See note to figure 2.8 for further explanations. Excluding countries with an initial poverty headcount rate below 3 percent. 100 UNDERSTANDING SHARED PROSPERITY headcount first on the growth of mean income or consumption for the population as a whole and then on the growth of mean income or con- sumption for the bottom 40 percent of the population. This reveals that the slopes of the two lines are not statistically different from each other, but the goodness of fit as measured by the R-squared is higher for growth of the bottom 40 percent than for growth of the population as a whole. This suggests that growth in shared prosperity explains more of the variation in observed changes in poverty.10 Although this evidence is not causal, it does suggest that boosting shared prosperity could be particularly relevant for poverty reduction. Finally, to explore the relationship between shared prosperity and pov- erty reduction explicitly, this section ends by posing the following question: how would global poverty in 2030 change if countries performed well by boosting shared prosperity? As chapter 1 shows, even the most aspirational growth scenario (assuming each country will grow at the rate of its best growth episode during the past 20 years) does not manage to bring poverty down to the goal of 3 percent globally. This section revisits the simulations by extending the projections to include scenarios where both growth and the distribution of growth change. Table 2.1 presents the results. As a benchmark, the first column in table 2.1 assumes that each country grows at its annualized growth rate over the last 10 years (with no changes in the distribution of growth). In this scenario, the extreme poverty head- count among developing countries by 2030 would be 5.6 percent, while the global headcount would be 4.8 percent. Now, consider a scenario where each country again grows at its annual- ized growth rate over the last 10 years, but this time the bottom percentiles grow faster than those at the top. For this simulation, in each country, the same growth incidence curve is imposed using the respective regional best performers in terms of relative pro-poorness while maintaining the country-specific annualized growth rate projection fi xed. The incidence curves are chosen based on the difference between growth of the bottom 40 percent relative to the mean while trying to maximize comparability of data and periods. For this simulation, the best performers in their respec- tive regions are Brazil (2001 to 2009), Jordan (2003 to 2011), Rwanda (2000 to 2011), Sri Lanka (2002 to 2010), and Thailand (2000 to 2010). In all these spells, the bottom 40 percent grew faster than the mean, with the difference ranging between 0.34 percentage points in Rwanda and 3.20 percentage points in Brazil.11 The growth incidence of these countries is then imposed on the 44 countries used for the exercise up to 2030 while 101 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Table 2.1 Reaching the 2030 extreme poverty goal in a world of higher shared prosperity (percent of population below $1.25-a-day poverty line) Scenario 3. Scenario 2. 10-year growth rate with Scenario 1. 10-year growth income distributions of 10-year growth rate rate with income the best regional performers with actual income distributions of the and additional 1% growth Region distributions best regional performers in Sub-Saharan Africa East Asia and the Pacific 0.3 0.2 0.2 Europe and Central Asia 0.0 0.0 0.0 Latin America and the Caribbean 3.2 1.5 1.5 Middle East and North Africa 1.1 1.1 1.1 South Asia 1.6 0.3 0.3 Sub-Saharan Africa 23.9 22.6 16.5 Total developing world 5.6 4.7 3.6 World 4.8 4.1 3.1 Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: The three scenarios present year 2030 projections of headcount poverty rates using the $1.25-a-day international poverty line. Scenario 1 assumes that each country grows at its historical 10-year national accounts growth rate. Scenarios 2 and 3 assume distributional changes in all 44 countries that are projected to have more than one million poor people in 2030 when distributions remain unchanged. Scenario 2 assumes that countries will grow at the same historical 10-year national accounts growth rate, but that the bottom of the income (or consumption) distribution benefits relatively more from growth than the top of the distribution. The incidence of growth in each country is determined by the observed past growth incidence curve of a regional leader over the past 10 years in terms of growth of the bottom 40 percent relative to mean growth. Poverty headcounts in 2030 are then projected by combining incidence curves with country-specific regional growth rates. Scenario 3 adds the assumption that national accounts growth in Sub-Saharan Africa is 1 percentage point faster than over the past 10 years. Also see endnote 11. maintaining country-specific overall projected annualized growth rates. To simplify the exercise, rather than considering all developing countries, the estimation focuses only on changes in the 44 countries that contribute the most to the global headcount in 2030 and that have populations of more than one million poor people. The results can thus be seen as a lower bound to the distributional effect.12 In this scenario of higher shared prosperity, the extreme poverty headcount for developing countries is decreased by an additional per- centage point to 4.7 percent, while the global headcount decreases to 4.1 percent. Although these outcomes are significantly better, distributional shifts do not close the gap to reach the 3 percent goal compared with the 102 UNDERSTANDING SHARED PROSPERITY growth-alone scenario. A primary reason seems to be that although this scenario contributes to faster poverty reduction in middle-income coun- tries, particularly in Latin America and the Caribbean and South Asia, it does not help to decrease poverty in low-income countries, particularly in Sub-Saharan Africa. This partly reflects that nature of the exercise: the dis- tribution of income (consumption) of the regional leader in Sub-Saharan Africa (Rwanda) is significantly less progressive than that of the regional leader in Latin America and the Caribbean (Brazil), thus the simulated impact of the exercise for countries in Sub-Saharan Africa is smaller. In addition, countries in Sub-Saharan Africa experienced zero or even nega- tive per capita growth rates during the period the simulation covers. For these countries, the potential impact on poverty reduction from distribut- ing a small amount of growth more widely is clearly limited. To demonstrate the particular sensitivity of poverty reduction in Sub- Saharan Africa to overall growth, a third scenario illustrates how much additional impact on poverty can be achieved by twinning improved growth performance with a more pro-poor incidence of growth. Table 2.1 therefore presents an additional scenario, which adds to scenario 2 an extra assump- tion of 1 percentage point additional growth in Sub-Saharan African coun- tries. In this scenario, the global headcount falls to the vicinity of 3 percent of the world’s population; the global extreme poverty goal is reached. Figure 2.12 provides two additional simulations. Each country’s initial mean is projected to grow between 2011 and 2030 following the histori- cal annualized growth rate over the last 10 years (similar to simulation 1 in table 2.1). The two new simulations add the twist that the bottom 40 percent of the distribution grows faster than the top 60 percent of the distribution, while preserving the growth in the mean. If the growth rate of the bottom 40 percent is assumed to be 1 percent- age point greater than the growth rate of the mean, the global extreme poverty headcount would decrease by an additional 1.1 percentage points relative to the scenario where all incomes grow at the annualized growth rate over the last 10 years (figure 2.12, red line).13 With even higher shared prosperity of 2 percentage points above the mean, the global 3 percent poverty target would be reached by 2028. Thus, twinning growth with improvements in shared prosperity can make the difference in closing the gap to meet the global poverty target. Regionally, however, the picture is not that simple. While shared pros- perity can have a significant effect on the global headcount, particularly in a context where growth is low, Sub-Saharan Africa stands out as an 103 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 2.12 Twinning growth and shared prosperity to reach the 2030 extreme poverty goal 18 15 Global poverty headcount (percent) 14 12 10 8 6 4 3% target 2 0 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 10-year historical growth rate 10-year historical growth rate, plus 1 percentage point faster growth for bottom 40 10-year historical growth rate, plus 2 percentage point faster growth for bottom 40 Source: Based on data from the World Bank PovcalNet database (accessed August 2014). Note: The three scenarios present projections from 2011 to 2030 of the global poverty headcount using the $1.25-a-day international poverty line. Scenario 1 (blue line) assumes that each country grows at its historical 10-year national accounts growth rate. Past national accounts growth rates are calculated as the annualized growth rate of real GDP per capita (countries in Sub-Saharan Africa) or the annualized growth rate of household final consumption expenditure per capita (all other countries) over the period 2000–10 (see also notes to table 1.4 in chapter 1). Scenario 2 (red line) assumes that the bottom 40 percent have an annualized 19-year growth rate (between 2011 and 2030) that is 1 percentage point greater than the annualized growth in the mean over this period. Scenario 3 (orange line) repeats scenario 2 but with a growth differential of 2 percentage points. exception. Distributional shifts in this region’s countries would contribute relatively less to reducing the regional headcount. Chapter 1 shows that projecting the annualized growth from the 2000s would bring the propor- tion of the poor in Sub-Saharan Africa from 46.8 percent to 23.9 percent. If the bottom 40 percent grows 1 or 2 percentage points faster than the mean, this would reduce the regional headcount by an additional 4 or 8 percentage points, respectively. Although this effect is greater than the simulation based on regional best performers, it still suggests that extreme poverty rates would remain very high in Sub-Saharan Africa under all of the simulations considered. Overall, these results point to the inherent complementarities between the World Bank’s two goals whereby a combination of substantial growth 104 UNDERSTANDING SHARED PROSPERITY and comparatively higher growth rates of the income of the bottom 40 percent could significantly improve the chances of reaching the global headcount goal of 3 percent. Key conclusions on shared prosperity This chapter has explored the conceptual and empirical underpinnings of the World Bank’s new shared prosperity goal. Five take-away messages stand out. First, the shared prosperity goal seeks to increase sensitivity to distributional issues, shifting the common understanding of development progress away from per capita income and emphasizing that good growth should benefit not just the wealthiest, but the least well-off in society as well. Although distributional issues have been part of the development debate for decades, this is the first time that an indicator with close links to inequality of outcomes (incomes) has become a benchmark of develop- ment progress. Second, unlike the global poverty goal, boosting shared prosperity is a country-specific goal with no explicit target. This has an inherent simplic- ity in that countries can track their own performance and have different aspirational targets. Still, interpreting performance at the country level may not be straightforward in the absence of a clear standard of what constitutes “good” performance. Looking at trends in shared prosperity over time and comparing the average income growth of the bottom 40 percent with the rest of the population over time provide two simple ways of assessing performance. Third, the goal is relevant for poor countries. In low- and lower-middle- income countries, there will likely be significant overlap between those living in absolute poverty and the bottom 40 percent of the population. Boosting shared prosperity should therefore reinforce poverty reduction efforts in these countries. In doing so, the added emphasis to improve the living standards of the poor may be instrumental for achieving the ambi- tious goal of ending global poverty. As the evidence in chapter 1 suggests, growth in GDP in the most optimistic scenarios will not be sufficient to reach the 3 percent global poverty target unless a structural shift happens. This chapter indicates that a scenario with higher shared prosperity can facilitate reaching the extreme poverty goal. At the same time, the chapter makes clear that transformational changes in growth and shared prosper- ity are needed to reach the poverty goal: repeating historical performance is not enough. 105 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Fourth, the goal is also relevant for richer countries. A substantial proportion of the bottom 40 percent of the population in upper-middle- income countries is likely to be nonpoor by the $1.25 per day global standard. The shared prosperity focus in these settings will bring atten- tion to those not covered by poverty policies but who might otherwise be relatively left behind and highly vulnerable to reentering poverty if shocks affect them. Finally, many challenges in measuring and tracking shared prosperity remain. Monitoring the goal requires high-quality and frequent household survey data. Issues of comparability across surveys—relevant also for consis- tently tracking global poverty—become more pronounced when looking at the performance of shared prosperity. Given that estimates of shared pros- perity are sensitive to the time period and intervals used, comparisons over time should be treated with caution. It is advisable to take advantage of all information (for example, calculating moving averages in shared prosperity when feasible). Cross-country comparisons should also be treated with cau- tion. The frequency and quality of household survey data across the world are heterogeneous, with some countries having a considerable way to go in producing the consistent and reliable data needed to track shared prosperity over time. Supporting these efforts is an important policy priority. Notes 1. This chapter uses income and consumption interchangeably, although, as chapter 1 notes, consumption is the preferred choice. 2. The official World Bank approach endorses use of the mean of the bottom 40 percent of the population, but it is worth noting that IDA17’s Results Measurement Framework has instead adopted the median income growth rate of the bottom 40 percent as an indicator of progress (World Bank 2014a). 3. Chapter 3 provides a more in-depth discussion of these issues. A discussion of the axiomatic requirements of welfare functions can also be found in Foster and others (2013) and Campbell and Kelly (2002). 4. The concept of people living on between $4 and $10 a day being considered vulnerable is based on evidence that a considerable share of households above a given poverty line is usually vulnerable to falling below that line over time. See López-Calva and Ortiz-Juarez (2014); Ferreira and others (2012); and Birdsall, Lustig, and Meyer (2014). 5. The World Bank regions have begun producing operational reports that provide more detailed profiles of the bottom 40 percent in each country (see World Bank 2014b, Rama and others 2014, or Bussolo and López-Calva 2014). 106 UNDERSTANDING SHARED PROSPERITY 6. See also Beegle and others (2014) for a discussion of the potential trade-offs. 7. For the technically curious, this is a case where the weak transfer axiom discussed earlier in this chapter comes into play. 8. Official estimates of shared prosperity will be published in the forthcoming Global Monitoring Report. 9. Chenery and others (1974) discuss these issues in the context of overall growth. See Rosenblatt and McGavock (2013) for a more recent discussion. 10. The results hold when the regressions are weighted by the initial poverty headcount. 11. These are not based on the official shared prosperity estimates. 12. Three other caveats should be noted. First, the estimates are mostly based on consumption data, which typically present lower inequality values because savings uniformly increase with income. The consumption share of the top earners is not as high as their income share. The effects of distributional changes in income could therefore be considerably higher than shown for consumption. Second, the estimates are very sensitive to missing data on top earners, as the latter contribute significantly—and proportionally—more to average income, which directly biases upward the difference between growth in the bottom 40 percent and that of the mean. Finally, since distributive government social expenditures such as in health and education are not always captured in consumption surveys, important government inequality- reducing tools are not captured in this framework. 13. To ensure that the mean growth rate is equal to the historical 10-year average, the growth rate of the top 60 percent is adjusted downward to compensate for the faster growth of the bottom 40 percent. In some cases, the differen- tial growth rates lead to a reranking of incomes across the 40th percentile of the distribution. To account for the reranking, the growth rate differential between the bottom 40 percent and the mean in each year is assumed to be somewhat greater than 1 percentage point, so that the measured (ex post) annual growth rate differential between 2011 and 2030 is equal to exactly 1 percentage point. References Basu, Kaushik. 2001. “On the Goals of Development.” In Frontiers of Development Economics: The Future in Perspective, edited by Gerald M. Meier and Joseph E. Stiglitz, 61–86. Washington, DC, and New York: World Bank and Oxford University Press. ———. 2006. “Globalization, Poverty, and Inequality: What Is the Relationship? What Can Be Done?” World Development 34 (8): 1361–73. doi:10.1016/j. worlddev.2005.10.009. Beegle, Kathleen, Pedro Olinto, Carlos E. Sobrado, Hiroki Uematsu, and Yeon Soo Kim. 2014. “Ending Extreme Poverty and Promoting Shared Prosperity: Could There Be Trade-Offs Between These Two Goals?” Inequality in Focus 3 (1): 1–6. 107 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Birdsall, Nancy, Nora Lustig, and Christian J. Meyer. 2014. “The Strugglers: The New Poor in Latin America?” World Development 60 (August): 132–46. doi:10.1016/j.worlddev.2014.03.019. Bussolo, Maurizio, and López-Calva, Luis F. 2014. Shared Prosperity: Paving the Way in Europe and Central Asia. Washington, DC: World Bank Group. Campbell, Donald E., and Jerry S. Kelly. 2002. “Impossibility Theorems in the Arrovian Framework.” In Handbook of Social Choice and Welfare, 1: 35–94. Handbooks in Economics 19. Amsterdam: North-Holland. Chenery, Hollis B., Montek S. Ahluwalia, C. L. G. Bell, John H. Duloy, and Richard Jolly. 1974. Redistribution with Growth: Policies to Improve Income Distribution in Developing Countries in the Context of Economic Growth. London: Oxford University Press. Deaton, Angus. 1997. The Analysis of Household Surveys: A Microeconometric Approach to Development Policy. Baltimore, MD: Johns Hopkins University Press. Deaton, Angus, and Salman Zaidi. 2002. “Guidelines for Constructing Consump- tion Aggregates for Welfare Analysis.” Living Standards Measurement Study Working Paper 135, World Bank, Washington, DC. Dollar, David, Tatjana Kleineberg, and Aart Kraay. 2013. “Growth Still Is Good for the Poor.” Policy Research Working Paper 6568, World Bank, Washington, DC. Essama-Nssah, B., and Peter J. Lambert. 2009. “Measuring Pro-Poorness: A Unifying Approach with New Results.” Review of Income and Wealth 55 (3): 752–78. doi:10.1111/j.1475-4991.2009.00335.x. Ferreira, Francisco H. G., Julian Messina, Jamele Rigolini, Luis-Felipe López- Calva, Maria Ana Lugo, and Renos Vakis. 2012. Economic Mobility and the Rise of the Latin American Middle Class. Washington, DC: World Bank. Foster, James E., Suman Seth, Michael Lokshin, and Zurab Sajaia. 2013. A Unified Approach to Measuring Poverty and Inequality: Theory and Practice; Streamlined Analysis with ADePT Software. Washington, DC: World Bank. Kakwani, Nanak Chand, and Ernesto M. Pernia. 2000. “What Is Pro-Poor Growth?” Asian Development Review 18 (1): 1–16. Klasen, Stephan. 2004. “In Search of the Holy Grail: How to Achieve Pro- Poor Growth?” In Proceedings from the Annual World Bank Conference on Development Economics—Europe, 2003: Toward Pro-Poor Policies: Aid, Institutions, and Globalization, edited by Bertil Tungodden, Nicholas Stern, and Ivar Kolstad, 63–93. Washington, DC, and New York: World Bank and Oxford University Press. ———. 2008. “Economic Growth and Poverty Reduction: Measurement Issues Using Income and Non-Income Indicators.” World Development 36 (3): 420–45. doi:10.1016/j.worlddev.2007.03.008. Kraay, Aart. 2006. “When Is Growth Pro-Poor? Evidence from a Panel of Countries.” Journal of Development Economics 80 (1): 198–227. doi:10.1016/j. jdeveco.2005.02.004. Lakner, Christoph, and Branko Milanovic. 2013. “Global Income Distribution: From the Fall of the Berlin Wall to the Great Recession.” Policy Research Working Paper 6719, World Bank, Washington, DC. 108 UNDERSTANDING SHARED PROSPERITY López-Calva, Luis F., and Eduardo Ortiz-Juarez. 2014. “A Vulnerability Approach to the Definition of the Middle Class.” Journal of Economic Inequality: 12(21), 23–47. McNamara, Robert S. 1972. Annual Address to the 1972 Annual Meetings of the Boards of Governors . Washington, DC: International Bank for Reconstruction and Development, International Finance Corporation, International Development Association. Narayan, Ambar, Jaime Saavedra-Chanduvi, and Sailesh Tiwari. 2013. “Shared Prosperity. Links to Growth, Inequality and Inequality of Opportunity.” Policy Research Working Paper 6649, World Bank, Washington, DC. Narayan, Ambar, Nobuo Yoshida, and Johan A. Mistiaen. 2013. “Computing Shared Prosperity Indicator for a Global Database.” Unpublished manuscript, World Bank, Washington, DC. Negre, Mario. 2010. “Concepts and Operationalization of Pro-Poor Growth.” WIDER Working Paper 2010/47, United Nations University, Helsinki. Palma, José Gabriel. 2011. “Homogeneous Middles vs. Heterogeneous Tails, and the End of the ‘Inverted-U’: It’s All About the Share of the Rich.” Development and Change 42 (1): 87–153. doi:10.1111/j.1467-7660.2011.01694.x. Rama, Martin, Tara Beteille, Yue Li, John Newman, and Pradeep Mitra. 2014. Inequality in South Asia. Washington, DC: World Bank Group. Ravallion, Martin. 1994. Poverty Comparisons . Fundamentals of Pure and Applied Economics 56. Chur, Switzerland: Harwood Academic Publishers. Ravallion, Martin, and Shaohua Chen. 2003. “Measuring Pro-Poor Growth.” Economics Letters 78 (1): 93–99. doi:10.1016/S0165-1765(02)00205-7. Ravallion, Martin, Shaohua Chen, and Prem Sangraula. 2009. “Dollar a Day Revisited.” The World Bank Economic Review 23 (2): 163–84. doi:10.1093/ wber/lhp007. Rawls, John. 1971. A Theory of Justice. Cambridge, MA: Belknap Press of Harvard University Press. Rosenblatt, David, and Tamara J. McGavock. 2013. “A Note on the Simple Algebra of the Shared Prosperity Indicator.” Policy Research Working Paper 6645, World Bank, Washington, DC. Sen, Amartya. 1983. “Development: Which Way Now?” The Economic Journal 93 (372): 745–62. doi:10.2307/2232744. ———. 1985. Commodities and Capabilities. Amsterdam: North-Holland. ———. 1999. Development as Freedom. Oxford: Oxford University Press. Skoufias, Emmanuel, Sailesh Tiwari, and Akhmad Rizal Shidiq. 2014. “Sharing Prosperity: The Poverty, Growth and Equity Nexus in Thailand.” Policy Research Working Paper, World Bank, Washington, DC. Slesnick, Daniel T. 1998. “Empirical Approaches to the Measurement of Wel- fare.” Journal of Economic Literature 36 (4): 2108–65. Son, Hyun Hwa. 2004. “A Note on Pro-Poor Growth.” Economics Letters 82 (3): 307–14. doi:10.1016/j.econlet.2003.08.003. Subramanian, Subbu. 2011. “‘Inclusive Development’ and the Quintile Income Statistic.” Economic and Political Weekly XLVI (4): 69–72. van der Weide, Roy, and Branko Milanovic. 2014. “Inequality Is Bad for Growth of the Poor (But Not for That of the Rich).” Policy Research Working Paper 6963, World Bank, Washington, DC. 109 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY World Bank. 1990. World Development Report 1990: Poverty. New York: Oxford University Press. ———. 2014a. Report from the Executive Directors of the International Development Association to the Board of Governors, Additions to IDA Resources: Seventeenth Replenishment—IDA17: Maximizing Development Impact. World Bank Group, Washington, DC. ———. 2014b. Social Gains in the Balance: A Fiscal Policy Challenge for Latin America and the Caribbean. Washington, DC: World Bank. 110 CHAPTER THREE The Twin Goals in a Broader Context The previous two chapters have discussed in detail the conceptual under- pinnings and data requirements of the World Bank’s twin goals of eliminat- ing global extreme poverty and promoting shared prosperity. This chapter provides further conceptual and empirical perspective on the twin goals by setting them in a broader context. While the twin goals set by the World Bank imply a particular set of institutional preferences or priorities across individuals, this does not mean that these should be the only valid priorities for all development partners. National governments, other aid donors, or any other group might choose to emphasize other priorities that best reflect their particular objectives. The scope for such differences in priorities is perhaps clearest in the context of poverty measurement. Although the World Bank has placed emphasis on the fraction of people living below the global poverty line as a global objective, national governments attach priority to poverty thresh- olds that are relevant to their particular countries—as evidenced by large differences across countries in national poverty lines. Similarly, when ana- lyzing poverty in a particular country, best practice quickly goes beyond the headcount measure of poverty to consider other poverty measures that capture the severity as well as the incidence of poverty. The scope for different priorities can also be seen in the context of the shared prosperity measure, which ascribes particular importance to the share of total income that goes to the poorest 40 percent of people in a country. Yet the income share of the bottom 40 percent is just one of many measures of how equitably or inequitably income is distributed across individuals in a country. Different inequality measures imply very different priorities over individuals at different points in the income dis- tribution. For some purposes, a country might choose to prioritize those 111 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY in the bottom 20 percent, or even the bottom 90 percent, rather than the bottom 40 percent. Or it might choose to evaluate the benefit of a policy intervention in terms of its effects on individuals throughout the entire income distribution, with different weights for people at different income levels. This chapter uses social welfare functions as a tool of analysis to set the twin goals in this broader context.1 Economists have long used social welfare functions to capture societal preferences over how income is dis- tributed across individuals in a society (box 3.1). The twin goals set by Box 3.1 Social welfare functions articulate priorities across individuals The World Bank’s twin goals of eliminating • The second ingredient consists of the welfare extreme poverty and boosting shared prosperity weights that a given social welfare function are two particular cases in a large class of social assigns to different percentiles of the income welfare functions, which economists have long used distribution. The blue line in figure B3.1.1 shows to represent preferences over how income is distrib- a possible set of such welfare weights across per- uted across individuals. centiles, based on the income of each percentile, Two key ingredients are required for this that is, w (y (p)). In this case, social preferences analysis: assign higher weights to the poor and lower weights to the rich. In other words, the welfare • The first ingredient is the distribution of income weights are downward sloping. These weights across a population of interest (typically, a coun- show the importance that a society assigns to try). A convenient way of representing this is to individuals at different points in the income use a quantile function , which gives the income distribution. These in turn can be used to evalu- level y ( p) corresponding to each percentile ate policies. For example, the downward-sloping p ∈ [0,100] of the income distribution in that welfare weights shown in the figure imply that country. The quantile function is the inverse of this society would be in favor of policies that the cumulative distribution of income, that redistribute income from the rich to the poor. is, y (p) = F –1(p), where F (y) is the distribution The extent of desired redistribution depends on function of income and F –1(p) is the inverse of the relative weights assigned to the poor versus this function. The orange line in figure B3.1.1 the rich—the greater the weights assigned to the shows a quantile function, which is upward slop- poor relative to the rich, the more redistribution ing since poorer percentiles have lower income is desired. levels while richer percentiles have higher income levels. The steeper the quantile function is, the Based on these two ingredients, the social wel- greater are the income gaps between the rich fare function is the average of the welfare weights, (those to the right in the graph) and the poor represented as the shaded area below the blue line (those to the left). in figure B3.1.1—formally, W = ∫ w (y (p))dp. The (continued) 112 THE TWIN GOALS IN A BROADER CONTEXT Box 3.1 Continued Figure B3.1.1 Income distributions and social welfare functions Income of percentile p, y ( p) Income Welfare-weighted income of percentile p, w ( y ( p)) Social welfare, W = ͐ w ( y ( p))dp Percentile of income distribution, p social welfare function illustrated in the fi gure above the poverty line. This is because changes in assigns greater weights to those at lower percen- the incomes of anyone above the poverty line do not tiles of the income distribution. However, other affect the poverty measure. The following section social welfare functions may assign greater weight then discusses social welfare functions that do not to the rich. For example, average income is itself an rely on a poverty line. Social welfare functions that example of a social welfare function with increas- value equality will assign higher weights to poorer ing rather than decreasing weights. Since average percentiles. However, they may or may not assign income weights the incomes of everyone equally, positive weights throughout the entire distribu- it naturally assigns greater weight to those in richer tion. For example, the World Bank’s second goal is percentiles of the income distribution, since richer based on average incomes in the bottom 40 percent, percentiles have higher incomes. Intuitively, a 1 which implies zero welfare weights for those above percent increase in the income of a rich person the 40th percentile of the income distribution. raises average income by more than a 1 percent Finally, it is worth noting that these social welfare increase in the income of a poor person does, since functions are chosen because of their links to stan- the absolute change in income of the rich person dard poverty and inequality measures. In contrast, is larger. many theoretical models define welfare in terms The next section of this chapter discusses social of the present value of the lifetime utility of an welfare functions featuring a given poverty line, agent, which will depend on all the ingredients in such as the headcount measure of poverty on which the particular model under consideration. The final the World Bank’s first goal is based. Such social wel- section of this chapter discusses how such measures fare functions assign positive weights to those below can be used to aggregate across different dimensions the poverty line, but imply zero weight for those of well-being in a theory-consistent way. 113 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY the World Bank can be thought of as two particular cases of such social welfare functions and valuable insights about the twin goals can be learned by considering them in this broader context. Specifically, setting the twin goals in the context of a broader class of social welfare functions helps to clarify the precise institutional priorities implied by the goals and contrasts them with other potential priorities implied by other social welfare func- tions. The chapter also provides empirical perspective on the twin goals by documenting trends in the relevant measures and comparing them with the trends implied by other social welfare functions. Welfare functions with poverty lines The World Bank’s global poverty goal is based on the headcount measure of poverty. As discussed in chapter 1, this consists of counting up the number of people below a specified poverty line and expressing the sum as a fraction of the total population. By simply counting up the poor, this measure has the virtue of clarity, a key quality to crystalize political trac- tion around goals. This clarity comes at a cost, however, since it provides no information on the well-being of those below the poverty line, beyond the fact that they are poor. The cases of Pakistan (in 2007) and Senegal (in 2005), shown in figure 3.1, provide a vivid illustration. In both coun- tries, the proportion of the population living on less than $2 per day was the same, at 60 percent. However, average consumption levels below the poverty line were substantially lower in Senegal than in Pakistan. In the case of Pakistan, average consumption of the poor fell 60 cents short of the poverty line, whereas the shortfall in Senegal was considerably larger, at 82 cents. Such differences in the distribution of income below the poverty line are not captured by the headcount measure of poverty, and so other distri- butionally sensitive measures are commonly used to capture how “deep” or “severe” poverty is. The Foster-Greer-Thorbecke (FGT) class of poverty measures is the most commonly used because of its straightforward inter- pretation.2 The FGT class weights poor people according to their distance from the poverty line. Specifically, the weight assigned to each poor person is the gap between their income and the poverty line, expressed as a frac- tion of the poverty line, and raised to an exponent. When the exponent is zero, the index weights everyone below the poverty line equally, resulting in the headcount measure. When the exponent is one, the index is the 114 THE TWIN GOALS IN A BROADER CONTEXT Figure 3.1 The headcount provides an incomplete picture of well-being below the poverty line 10.0 Consumption expenditure per capita per day (2005 PPP$, log scale) $2-a-day poverty line 1.0 0.1 0 20 40 60 80 100 Percentile of consumption distribution Consumption of percentile p: Average consumption of poor: Pakistan Senegal Pakistan Senegal Source: Based on data from the World Bank PovcalNet database. Note: This graph shows the distribution of consumption (on the vertical axis) for Pakistan in 2007 (the blue line) and Senegal in 2005 (the orange line). The horizontal gray line shows the $2-a-day line. The red and blue dashed lines show average income of those below the poverty line in Pakistan and Senegal, respectively. The consumption distributions are lognormal approximations to the true distributions. PPP = purchasing power parity. average gap between the incomes of the poor and the poverty line, as a fraction of the poverty line. When the exponent is two, the measure is the squared poverty gap. By placing greater weight on those further below the poverty line, this measure also reflects inequality among the poor. Figure 3.2 summarizes the weights that members of the FGT class of poverty measures assign to individuals in different percentiles of the income distribution. The headcount measure weights everyone below the poverty line equally. In contrast, the poverty gap and the squared poverty gap are social welfare functions that place successively higher weights on the poorest. These weights have important implications for policy choices. For a fixed amount of resources, if reducing the headcount is taken as the primary goal, the most effective use of funds would imply focusing on the poor nearest to the poverty line, so that as many of them as possible go over the poverty threshold and hence decrease the poverty headcount. This may not be ideal, however, because it biases poverty reduction efforts away 115 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Figure 3.2 Welfare weights implied by different poverty measures Weight on percentile p in social welfare function 0 20 40 60 80 100 Percentile of consumption distribution Headcount Poverty gap Squared poverty gap Source: Based on data from the Bangladesh Household Income and Expenditure Survey (2010). Note: This graph plots the weights assigned by common poverty measures to individuals at different ⎛ z − y ( p)⎞ θ points in the income distribution. The weights are w ( y ( p)) = I y ( p )< z with q = 0 for the headcount, ⎝ z ⎠ q = 1 for the poverty gap, and q = 2 for the squared poverty gap. The weights have been normalized to sum to one, and are drawn for the observed distribution of household consumption expenditures in Bangladesh in 2010. The poverty line is set at 1,487 Bangladesh taka per month, which generates a headcount in 2010 of 31.5 percent. from the poorest among the poor. Means-tested antipoverty programs can be thought of as a way of avoiding this problem. For example, food stamps (vouchers to be used for food purchases by the poor) are an important part of the social safety net in the United States. Food stamp benefits are means tested, in the sense that the value of the benefit declines as the incomes of the poor increase. Since this program provides a greater benefit to the poorest, it has a greater proportional impact on the poverty gap and the squared poverty gap than it does on the poverty headcount (Jolliffe and others 2005). Figure 3.3 provides a systematic look at the practical consequences of these different approaches to weighting individuals below the poverty line, drawing on the most recent household survey available in the PovcalNet database. The figure shows recent data for 81 countries for which the $1.25 a day headcount measure of poverty constitutes at least 5 percent of the population. Panels a and b, respectively, graph the headcount and the poverty gap (on the vertical axis) against the logarithm of the household 116 THE TWIN GOALS IN A BROADER CONTEXT Figure 3.3 Different poverty measures fall with income, but tell different stories a. Headcount versus income b. Poverty gap versus income Poverty headcount at $1.25 a day (percent) 50 80 Poverty gap at $1.25 a day (percent) 40 60 30 40 20 20 10 0 0 0 5 10 15 0 5 10 15 Mean income or consumption per capita Mean income or consumption per capita per day (2005 PPP$) per day (2005 PPP$) c. Poverty gap versus headcount d. Squared poverty gap versus headcount 40 Squared poverty gap at $1.25 a day (percent) 50 Poverty gap at $1.25 a day (percent) 40 30 30 20 20 10 10 0 0 0 20 40 60 80 0 20 40 60 80 Poverty headcount at $1.25 a day (percent) Poverty headcount at $1.25 a day (percent) Source: Based on data from the World Bank PovcalNet database. Note: Panels a and b graph the headcount and poverty gap against household survey mean income or consumption. Each data point corresponds to the most recently available survey in PovcalNet (as of July 2014) for a country. The sample is restricted to countries where the headcount ratio is greater than 5 percent of the population. PPP = purchasing power parity. 117 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY survey mean (on the horizontal axis). Both panels show a strong nega- tive relationship: across countries, both measures of poverty fall sharply as average incomes increase. However, there is also substantial variation around this average relationship. For example, countries such as Ethiopia and Togo, in 2011, have similar levels of survey mean consumption, just below $2 per day. Yet the $1.25 a day headcount measure of poverty is 37 percent in Ethiopia but 52 percent in Togo. This variation in poverty at a given level of average living standards illustrates the value of using social welfare functions that assign greater weight to the poor than to the rich. As discussed in box 3.1, average income corresponds to a social welfare function that weights the rich more heavily than the poor. Judged in terms of average incomes, Ethiopia and Togo are equally well-off. Yet by look- ing at poverty measures that assign greater weight to the poor than the rich, differences in social welfare across countries at a given income level become apparent. Panels c and d of figure 3.3 illustrate the extent to which differences in weights applied to people below the poverty line matter in practice. In both panels, the horizontal axis corresponds to the headcount measure of poverty (which weights everyone below the poverty line equally) and the vertical axes correspond to the poverty gap and the squared poverty gap, respectively. Overall, the more bottom-sensitive poverty measures track the simple headcount measure fairly closely across countries. However, as noted in the discussion of figure 3.1, in some cases countries with the same headcount level of poverty can have quite different poverty gaps or squared poverty gaps. Where the poverty line is drawn also has important implications for how individuals are weighted. This is because these poverty measures assign a weight of zero to those above the poverty line, indicating that the measures ignore the nonpoor entirely. This feature of standard poverty measures is uncomfortable, particularly when one considers those just above the poverty line. Not only are those just above the poverty line very similar in terms of their income or consumption to those just below the poverty line, but they also likely face a high risk of falling back into poverty. This, in turn, makes distinctions between those “just above” and “just below” a fixed poverty line somewhat arbitrary. Recognizing this, several studies have proposed various definitions of “vulnerability” to poverty, which seek to convey the sense that individuals currently above the poverty line face nonnegligible risks of falling back into poverty.3 For example, López-Calva and Ortiz-Juarez (2011) define a 118 THE TWIN GOALS IN A BROADER CONTEXT vulnerability zone just above the poverty line. This zone is defined in terms of an income level above the poverty line where the risk of falling back into poverty is 10 percent or more over a five-year interval. This higher thresh- old income level is estimated with panel data on transitions into and out of poverty in Chile, Mexico, and Peru. For national poverty lines between $4 and $5 in terms of purchasing power parity (PPP), a vulnerability zone between the poverty line and a higher threshold of $10 PPP a day is esti- mated. López-Calva and Ortiz-Juarez find that this seems to correspond well with subjective self-assessments of the risk of falling into poverty. This approach has the appeal of explicitly recognizing poverty dynamics: individuals who are not poor in one period may very well become poor in the next period. At the same time, however, the approach is somewhat asymmetric, since it does not recognize that those who are poor in one period might become nonpoor in the next period. A further drawback of a fi xed poverty line is that the accompanying poverty measures may become less and less relevant over time, as coun- tries grow richer and the fraction of the population below the poverty line falls. The same is true for a poverty line that is fixed across countries, as its relevance may be very different in countries at different income levels. For example, based on the World Bank’s global poverty line, 71 percent of the population of Malawi was poor in 2011, while only 5 percent of Brazil’s and 1 percent of Chile’s populations were poor. Setting a fixed poverty line across countries implies different social welfare functions across countries. In the case of the headcount below the global poverty line, a fi xed pov- erty line implies a particular concern for 71 percent of the population of Malawi but only 5 percent of the population of Brazil. Box 3.2 discusses how national poverty lines vary across countries, as well as proposals for a “weakly relative” international poverty line that varies with income levels. Beyond the poverty line: Social welfare functions that care about everyone The discussion in the previous section focused on social welfare functions featuring a fixed poverty line. A key feature of these measures is that they assign zero weight to individuals above the poverty line. This section turns to social welfare functions that do not distinguish between “the poor” and “the nonpoor” but rather assign weights throughout the income distribu- tion. The discussion here encompasses the second of the twin goals, in the 119 A MEASURED APPROACH TO ENDING POVERTY AND BOOSTING SHARED PROSPERITY Box 3.2 Where to draw the poverty line? It is a common occurrence to see eyebrows rise and Spain had poverty rates around 23 percent in when visitors from low per capita income countries 2012, slightly above the poverty rate in India based are in a high per capita income country and read on national poverty lines. Since these European local poverty statistics in the media. The visitors countries are vastly richer on average than India, find, to their surprise, that the figures are not far these differences come from the fact that the from the ones back at home. Based on national national poverty lines used are very different.a poverty lines, countries such as Greece, Romania, Figure B3.2.1, panel a, shows that in the poorest Figure B3.2.1 Weakly relative poverty lines, and global poverty based on weakly relative lines a. National poverty lines increase with the b. Global poverty falls more slowly if it is based level of development across countries on weakly relative poverty lines that increase with the level of development 70 300 National poverty line, per capita 60 per month (2005 PPP$) Headcount (percent) 200 50 40 100 30 20 0 10 3 4 5 6 7 1980 1985 1990 1995 2000 2005 2010 Log private consumption per capita Absolute poverty ($1.25 a day) per day (2005 PPP$) Weakly relative poverty (survey means) Weakly relative poverty (national accounts) Source: Adapted from Ravallion, Chen, and Sangraula (2009); Ravallion and Chen (2011); Chen and Ravallion (2013). Note: Solid line in panel a is a locally weighted regression smoother (lowess) with bandwidth = 0.8. PPP = purchasing power parity. (continued) sense that the average income of the bottom 40 percent of the population is not a social welfare function with a fi xed poverty line. Like measures with a fixed line, the shared prosperity measure assigns zero weight to indi- viduals in one part of the income distribution (those in the top 60 percent of the income distribution). However, this measure is just one particular 120 THE TWIN GOALS IN A BROADER CONTEXT Box 3.2 Continued countries in the world, poverty lines are not very line and project it forward through 2030 (figure different and are unrelated to per capita incomes. B3.2.1, panel b).e Global poverty is higher with After a certain point, however, national poverty the weakly relative line rather than the fixed global lines tend to increase more or less in line with the poverty line, since the weakly relative line implies level of national income. In European countries, higher poverty lines in richer countries. In addi- a person is generally considered poor if his or her tion, global poverty based on the weakly relative income falls below 60 percent of median income, so line declines more slowly over time, since poverty the poverty line automatically increases as median lines increase as countries’ incomes increase over income increases.b time. For the purpose of international comparisons, Sources: Figures adapted from Ravallion, Chen, and Ravallion, Chen, and Sangraula (2009) construct Sangraula (2009); Ravallion and Chen (2011); and Chen and the World Bank’s current global international pov- Ravallion (2013). erty line of $1.25 a day at 2005 purchasing power a. Government of India Planning Commission, in http://planningcommission.nic.in, visited on May 5, 2014; parity by taking the mean of the poverty lines in EUROSTAT, in http://appsso.eurostat.ec.europa.eu, visited the poorest 15 countries in terms of consumption on May 5, 2014. per capita.c They show that this is quite robust to b. See box 1.1 for a full discussion of how national poverty the choice of countries and that it is consistent with lines are set. the fact that poverty lines in the poorest countries c. See chapter 6 for a full discussion of how the international poverty line is set. do not vary much. Chen and Ravallion (2013) and d. In closely related work using the same data set of national Ravallion and Chen (2011) bring together the two poverty lines, Greb and others (2011) propose a weakly relative views by combining the global absolute poverty line poverty line that increases smoothly with the level of develop- of $1.25 for the poorest countries in the world with ment, rather than imposing a “kink” as Chen and Ravallion a “weakly relative poverty line” for richer countries, (2013) do. e. In the notation of box 3.1, the weakly relative poverty where the latter increases with per capita income line corresponds to a set of welfare weights given by w (y (p)) = (shown as the solid line in figure B3.2.1, panel a).d Iy(p)